{
  "schema_version": 2,
  "generated_at": "2026-09-12T03:34:48.297431Z",
  "metric": "Mean WS21",
  "scope": {
    "id": "G20",
    "stock": "GOOG",
    "market_dates": [
      "2026-01-02",
      "2026-01-05",
      "2026-01-06",
      "2026-01-07",
      "2026-01-08",
      "2026-01-09",
      "2026-01-12",
      "2026-01-13",
      "2026-01-14",
      "2026-01-15",
      "2026-01-16",
      "2026-01-20",
      "2026-01-21",
      "2026-01-22",
      "2026-01-23",
      "2026-01-26",
      "2026-01-27",
      "2026-01-28",
      "2026-01-29",
      "2026-01-30"
    ],
    "record_count": 357,
    "source_record_count": 376,
    "excluded_unknown_dates": 19,
    "source_date_count": 357,
    "estimated_count": 2,
    "verified_mean21_count": 25,
    "completion_date_count": 0,
    "date_policy": "source_recorded_date; unknown source dates excluded"
  },
  "source": {
    "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
    "snapshot_results_sha256": "1a5d36c7afbdf63530c4c81033a64feb7972b4551467534df3e7345ebb9bc045",
    "exporter_commit": "50d8aea86b7fd70fb2f61bbacad6f57b4bc4a257",
    "remote_repository_head_when_checked": "97a6b43efa2b302bd55323c2c8cd3676ef68c79a",
    "remote_directory": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912",
    "files": [
      {
        "filename": "evaluation_table_all_1290.json",
        "sha256": "69008b04a8db9df30e31747080cc62671a4ad6ba89abea6f5c2d22d02e5716dc",
        "bytes": 2060234
      },
      {
        "filename": "evaluation_table_all_1290.csv",
        "sha256": "3362e391cc43420db2d884f19d3f90d63e00036b198b2d04e7224e238c323b94",
        "bytes": 1411455
      },
      {
        "filename": "evaluation_table_all_1290.html",
        "sha256": "00d20b887749d60e50996867b31312e333dfc3924bcd62379168f13d60332df4",
        "bytes": 1208187
      },
      {
        "filename": "evaluation_table_audit.json",
        "sha256": "fb5884c7d59cd033c2b153422c36c3285e12544a25088bfdce51b29665e7ddb2",
        "bytes": 46820
      },
      {
        "filename": "g_classes_lowest.json",
        "sha256": "343f6dd7387190d5d1c2fb71769165eba6f39373a80ef5b5da1220357ba2d68e",
        "bytes": 80628
      },
      {
        "filename": "export_evaluation_table.py",
        "sha256": "6dd5207281cb854e5f084c42c427011be3b59501b700b0124e3cd2d1f75bdc56",
        "bytes": 21481
      },
      {
        "filename": "evaluation_table_G20_376.csv",
        "sha256": "f25661e6b7e5b9fa3a40424c40c69e81ed985d07c053175c5111392ab1f906e9",
        "bytes": 435909
      }
    ],
    "checked_on": "2026-09-12"
  },
  "results": [
    {
      "id": "bpe-field-v10000-seed137",
      "model": "Mamba3 field-v10000 seed 137",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.2035797823566927,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-field-v10000-seed137.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.29570671987804353,
        "ask_cancellation_ticks": 0.2952213596816538,
        "ask_volume": 0.025377803437866157,
        "ask_volume_touch": 0.049305858797188004,
        "bid_cancellation_depth": 0.31272252137790785,
        "bid_cancellation_ticks": 0.3128148633399748,
        "bid_volume": 0.034969728370571204,
        "bid_volume_touch": 0.03546933173396993,
        "limit_ask_order_depth": 0.21604485563253087,
        "limit_ask_order_ticks": 0.21611281091918977,
        "limit_bid_order_depth": 0.32966372888371387,
        "limit_bid_order_ticks": 0.3286332473577636,
        "log_inter_arrival_time": 0.06822448848933343,
        "log_time_to_cancel": 0.3681874115731461,
        "ofi": 0.2823666115411928,
        "ofi_down": 0.26533111181945207,
        "ofi_stay": 0.28322052995608993,
        "ofi_up": 0.2292819474470691,
        "orderbook_imbalance": 0.09360149965069724,
        "spread": 0.19076874850862674,
        "vol_per_min": 0.04215025109456558
      },
      "evaluation_key": "5f75c4ecc8ef23b9c18c6f3d7abfc5ec14e0111fe99f29d839d89fcaa385cb72",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2035797823566927,
      "g20_classification": {
        "id": "bpe-field-v10000-seed137",
        "class": "G20",
        "model_evaluation": "Mamba3 field-v10000 seed 137 | step 30000; pooled January validation",
        "wws21_mean": 0.2035797823566927,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/5",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 648,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.20357978235669266,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-field-v10000-seed137.json"
            }
          ],
          "evaluation_key": "5f75c4ecc8ef23b9c18c6f3d7abfc5ec14e0111fe99f29d839d89fcaa385cb72",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/field-v10000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/field-v10000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "39cb2d55d50c59fd2800fb0ec5b54511564cd5ac4842138b0b6434adc75fe673",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.20357978235669266,
        "feature_points": {
          "ask_cancellation_depth": 0.29570671987804353,
          "ask_cancellation_ticks": 0.2952213596816538,
          "ask_volume": 0.025377803437866157,
          "ask_volume_touch": 0.049305858797188004,
          "bid_cancellation_depth": 0.31272252137790785,
          "bid_cancellation_ticks": 0.3128148633399748,
          "bid_volume": 0.034969728370571204,
          "bid_volume_touch": 0.03546933173396993,
          "limit_ask_order_depth": 0.21604485563253087,
          "limit_ask_order_ticks": 0.21611281091918977,
          "limit_bid_order_depth": 0.32966372888371387,
          "limit_bid_order_ticks": 0.3286332473577636,
          "log_inter_arrival_time": 0.06822448848933343,
          "log_time_to_cancel": 0.3681874115731461,
          "ofi": 0.2823666115411928,
          "ofi_down": 0.26533111181945207,
          "ofi_stay": 0.28322052995608993,
          "ofi_up": 0.2292819474470691,
          "orderbook_imbalance": 0.09360149965069724,
          "spread": 0.19076874850862674,
          "vol_per_min": 0.04215025109456558
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/5",
            "source_value": 0.2035797823566927,
            "report_record_id": "bpe-field-v10000-seed137",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-field-v10000-seed137].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-field-v10000-seed137-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "field-v10000 seed 137 mean daily WS-21",
      "mean_ws21": 0.24991757388382072,
      "reported_value": 0.24991757388382072,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 137,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/field-v10000-seed137/scores/pooled/lobbench_summary.json",
      "upstream_sha256": "fe7279b2a7338d6d2fd0031c8458a00e69e6aa85bb0348f402cd81b029893ff2",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.24991757388382072,
      "g20_classification": {
        "id": "bpe-field-v10000-seed137-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "field-v10000 seed 137 mean daily WS-21",
        "wws21_mean": 0.24991757388382072,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 649,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/field-v10000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/field-v10000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "39cb2d55d50c59fd2800fb0ec5b54511564cd5ac4842138b0b6434adc75fe673",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.24991757388382072,
            "report_record_id": "bpe-field-v10000-seed137-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-field-v10000-seed42",
      "model": "Mamba3 field-v10000 seed 42",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.21012257070113385,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-field-v10000-seed42.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.30299063357296385,
        "ask_cancellation_ticks": 0.3024157134131251,
        "ask_volume": 0.04940702405563244,
        "ask_volume_touch": 0.05173908064887299,
        "bid_cancellation_depth": 0.30798193883903446,
        "bid_cancellation_ticks": 0.3080592129617409,
        "bid_volume": 0.060576017822842845,
        "bid_volume_touch": 0.04889004667926528,
        "limit_ask_order_depth": 0.2249718198858875,
        "limit_ask_order_ticks": 0.22504606008799605,
        "limit_bid_order_depth": 0.32604642110211807,
        "limit_bid_order_ticks": 0.32527109815506705,
        "log_inter_arrival_time": 0.06685534340762195,
        "log_time_to_cancel": 0.4214114937370019,
        "ofi": 0.2899598687839203,
        "ofi_down": 0.23323055557208963,
        "ofi_stay": 0.2919818249463656,
        "ofi_up": 0.22785229246680716,
        "orderbook_imbalance": 0.09767274644389254,
        "spread": 0.21779242428859835,
        "vol_per_min": 0.032422367852966746
      },
      "evaluation_key": "3e024a77bf201531df50e99c80a438707db30b1aef975a7e79088b72d87c5ed3",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.21012257070113385,
      "g20_classification": {
        "id": "bpe-field-v10000-seed42",
        "class": "G20",
        "model_evaluation": "Mamba3 field-v10000 seed 42 | step 30000; pooled January validation",
        "wws21_mean": 0.21012257070113385,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/3",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 650,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.21012257070113385,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-field-v10000-seed42.json"
            }
          ],
          "evaluation_key": "3e024a77bf201531df50e99c80a438707db30b1aef975a7e79088b72d87c5ed3",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "No checkpoint path found in accessible metadata. The canonical aggregate points to this seed-42 source summary, but both that summary and its artifact directory are permission denied. Step 30000 is recorded; a path has not been inferred from the model name.",
        "completeness": "unresolved",
        "expected_component_count": 1,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seed": 42,
        "source_metadata_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
        "source_metadata_availability": "permission_denied"
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.21012257070113385,
        "feature_points": {
          "ask_cancellation_depth": 0.30299063357296385,
          "ask_cancellation_ticks": 0.3024157134131251,
          "ask_volume": 0.04940702405563244,
          "ask_volume_touch": 0.05173908064887299,
          "bid_cancellation_depth": 0.30798193883903446,
          "bid_cancellation_ticks": 0.3080592129617409,
          "bid_volume": 0.060576017822842845,
          "bid_volume_touch": 0.04889004667926528,
          "limit_ask_order_depth": 0.2249718198858875,
          "limit_ask_order_ticks": 0.22504606008799605,
          "limit_bid_order_depth": 0.32604642110211807,
          "limit_bid_order_ticks": 0.32527109815506705,
          "log_inter_arrival_time": 0.06685534340762195,
          "log_time_to_cancel": 0.4214114937370019,
          "ofi": 0.2899598687839203,
          "ofi_down": 0.23323055557208963,
          "ofi_stay": 0.2919818249463656,
          "ofi_up": 0.22785229246680716,
          "orderbook_imbalance": 0.09767274644389254,
          "spread": 0.21779242428859835,
          "vol_per_min": 0.032422367852966746
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/3",
            "source_value": 0.21012257070113385,
            "report_record_id": "bpe-field-v10000-seed42",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-field-v10000-seed42].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-field-v10000-seed42-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "field-v10000 seed 42 mean daily WS-21",
      "mean_ws21": 0.26021491444453015,
      "reported_value": 0.26021491444453015,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 42,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
      "upstream_sha256": "aa204921d690be4037279d0c280c81897945b8161fae9f4b73956aff23106aac",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.26021491444453015,
      "g20_classification": {
        "id": "bpe-field-v10000-seed42-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "field-v10000 seed 42 mean daily WS-21",
        "wws21_mean": 0.26021491444453015,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 651,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "No checkpoint path found in accessible metadata. The canonical aggregate points to this seed-42 source summary, but both that summary and its artifact directory are permission denied. Step 30000 is recorded; a path has not been inferred from the model name.",
        "completeness": "unresolved",
        "expected_component_count": 1,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seed": 42,
        "source_metadata_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
        "source_metadata_availability": "permission_denied"
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.26021491444453015,
            "report_record_id": "bpe-field-v10000-seed42-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-field-v10000-seed5",
      "model": "Mamba3 field-v10000 seed 5",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.19695355508919907,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-field-v10000-seed5.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.2891675005920622,
        "ask_cancellation_ticks": 0.28871820411241805,
        "ask_volume": 0.048733458045536436,
        "ask_volume_touch": 0.053485652772652395,
        "bid_cancellation_depth": 0.2988862413616605,
        "bid_cancellation_ticks": 0.29852207846430395,
        "bid_volume": 0.05707455641111133,
        "bid_volume_touch": 0.041923963418505614,
        "limit_ask_order_depth": 0.20660037102397522,
        "limit_ask_order_ticks": 0.20654125372989277,
        "limit_bid_order_depth": 0.31907163760926216,
        "limit_bid_order_ticks": 0.3179061301620015,
        "log_inter_arrival_time": 0.03393549442782667,
        "log_time_to_cancel": 0.3748860164408866,
        "ofi": 0.24466302840393772,
        "ofi_down": 0.19897726851037445,
        "ofi_stay": 0.2454298239024867,
        "ofi_up": 0.22540724030707987,
        "orderbook_imbalance": 0.10221456010660213,
        "spread": 0.2341763708099482,
        "vol_per_min": 0.049703806260656166
      },
      "evaluation_key": "6cf629727a10ea5529de31b897eaf2295036496462896d169c6dfd57b4a2a9bf",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.19695355508919907,
      "g20_classification": {
        "id": "bpe-field-v10000-seed5",
        "class": "G20",
        "model_evaluation": "Mamba3 field-v10000 seed 5 | step 30000; pooled January validation",
        "wws21_mean": 0.19695355508919907,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/4",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 652,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.19695355508919904,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-field-v10000-seed5.json"
            }
          ],
          "evaluation_key": "6cf629727a10ea5529de31b897eaf2295036496462896d169c6dfd57b4a2a9bf",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/arm-specs/generation/field-v10000-seed5/arm-0.env",
            "evidence_line": 3,
            "evidence_text": "CHECKPOINT_ROOT=/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/campaign.json",
            "checkpoint_step_evidence_line": 9,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "f2c9f945cc8af203f2e0f02ddca8698779286396e711b15983947ff29f17a2f9",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19695355508919907,
        "feature_points": {
          "ask_cancellation_depth": 0.2891675005920622,
          "ask_cancellation_ticks": 0.28871820411241805,
          "ask_volume": 0.048733458045536436,
          "ask_volume_touch": 0.053485652772652395,
          "bid_cancellation_depth": 0.2988862413616605,
          "bid_cancellation_ticks": 0.29852207846430395,
          "bid_volume": 0.05707455641111133,
          "bid_volume_touch": 0.041923963418505614,
          "limit_ask_order_depth": 0.20660037102397522,
          "limit_ask_order_ticks": 0.20654125372989277,
          "limit_bid_order_depth": 0.31907163760926216,
          "limit_bid_order_ticks": 0.3179061301620015,
          "log_inter_arrival_time": 0.03393549442782667,
          "log_time_to_cancel": 0.3748860164408866,
          "ofi": 0.24466302840393772,
          "ofi_down": 0.19897726851037445,
          "ofi_stay": 0.2454298239024867,
          "ofi_up": 0.22540724030707987,
          "orderbook_imbalance": 0.10221456010660213,
          "spread": 0.2341763708099482,
          "vol_per_min": 0.049703806260656166
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/4",
            "source_value": 0.19695355508919907,
            "report_record_id": "bpe-field-v10000-seed5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-field-v10000-seed5].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-field-v10000-seed5-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "field-v10000 seed 5 mean daily WS-21",
      "mean_ws21": 0.2478299516965688,
      "reported_value": 0.2478299516965688,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 5,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/scores/field-v10000-seed5/pooled/lobbench_summary.json",
      "upstream_sha256": "d597c8595d17a6284ea15ff6daf713ca893d8cda81c99389c88c6e6c518302f9",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2478299516965688,
      "g20_classification": {
        "id": "bpe-field-v10000-seed5-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "field-v10000 seed 5 mean daily WS-21",
        "wws21_mean": 0.2478299516965688,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 653,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/arm-specs/generation/field-v10000-seed5/arm-0.env",
            "evidence_line": 3,
            "evidence_text": "CHECKPOINT_ROOT=/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/campaign.json",
            "checkpoint_step_evidence_line": 9,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "f2c9f945cc8af203f2e0f02ddca8698779286396e711b15983947ff29f17a2f9",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.2478299516965688,
            "report_record_id": "bpe-field-v10000-seed5-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-field-v10000-three-seed-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "field-v10000 three-seed mean of per-seed daily WS-21 means",
      "mean_ws21": 0.2526541466749732,
      "reported_value": 0.2526541466749732,
      "metric_name": "daily_ws21_mean.mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean across three independently trained seeds of each seed mean over daily WS-21 scores",
      "protocol_scope": "GOOG January validation; seeds 42, 5, 137; 20 trading days; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "reported_sample_sd": 0.006630493838927445,
      "notes": "exact source values in seed order [42,5,137]: [0.26021491444453015, 0.2478299516965688, 0.24991757388382072]",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2526541466749732,
      "g20_classification": {
        "id": "bpe-field-v10000-three-seed-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "field-v10000 three-seed mean of per-seed daily WS-21 means",
        "wws21_mean": 0.2526541466749732,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean across three independently trained seeds of each seed mean over daily WS-21 scores；GOOG January validation; seeds 42, 5, 137; 20 trading days; 3,136 windows/seed; generation seed 2026；exact source values in seed order [42,5,137]: [0.26021491444453015, 0.2478299516965688, 0.24991757388382072]；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 654,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/arm-specs/generation/field-v10000-seed5/arm-0.env",
            "evidence_line": 3,
            "evidence_text": "CHECKPOINT_ROOT=/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/campaign.json",
            "checkpoint_step_evidence_line": 9,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "f2c9f945cc8af203f2e0f02ddca8698779286396e711b15983947ff29f17a2f9",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/field-v10000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/field-v10000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "39cb2d55d50c59fd2800fb0ec5b54511564cd5ac4842138b0b6434adc75fe673",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run"
          }
        ],
        "note": "Partial checkpoint mapping: the aggregate contains seeds 5, 42 and 137. Recorded roots and selected step 30000 are resolved for seeds 5 and 137. The seed-42 source summary and artifact directory are permission denied, so its checkpoint path remains unresolved.",
        "completeness": "partial",
        "expected_component_count": 3,
        "resolved_component_count": 2,
        "unresolved_model_seeds": [
          42
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. The aggregate uses a common C250/G250 protocol across all three components. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.2526541466749732,
            "report_record_id": "bpe-field-v10000-three-seed-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-field-v10000-three-seed-pooled-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "field-v10000 three-seed mean pooled WS-21",
      "mean_ws21": 0.20355196938234185,
      "reported_value": 0.20355196938234185,
      "metric_name": "ws21.mean",
      "metric_kind": "unconditional_ws21",
      "aggregation_level": "mean across three independently scored model seeds",
      "protocol_scope": "GOOG January validation; pooled observations; seeds 42, 5, 137; 3,136 windows/seed; 20 trading days; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "reported_sample_sd": 0.006584551861592573,
      "notes": "exact source values in seed order [42,5,137]: [0.21012257070113385, 0.19695355508919907, 0.20357978235669266]",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.20355196938234185,
      "g20_classification": {
        "id": "bpe-field-v10000-three-seed-pooled-ws21-mean",
        "class": "G20",
        "model_evaluation": "field-v10000 three-seed mean pooled WS-21",
        "wws21_mean": 0.20355196938234185,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean across three independently scored model seeds；GOOG January validation; pooled observations; seeds 42, 5, 137; 3,136 windows/seed; 20 trading days; generation seed 2026；exact source values in seed order [42,5,137]: [0.21012257070113385, 0.19695355508919907, 0.20357978235669266]；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 655,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/arm-specs/generation/field-v10000-seed5/arm-0.env",
            "evidence_line": 3,
            "evidence_text": "CHECKPOINT_ROOT=/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/campaign.json",
            "checkpoint_step_evidence_line": 9,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "f2c9f945cc8af203f2e0f02ddca8698779286396e711b15983947ff29f17a2f9",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/field-v10000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/field-v10000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "39cb2d55d50c59fd2800fb0ec5b54511564cd5ac4842138b0b6434adc75fe673",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_field_v10000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0003/run"
          }
        ],
        "note": "Partial checkpoint mapping: the aggregate contains seeds 5, 42 and 137. Recorded roots and selected step 30000 are resolved for seeds 5 and 137. The seed-42 source summary and artifact directory are permission denied, so its checkpoint path remains unresolved.",
        "completeness": "partial",
        "expected_component_count": 3,
        "resolved_component_count": 2,
        "unresolved_model_seeds": [
          42
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. The aggregate uses a common C250/G250 protocol across all three components.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.20355196938234185,
            "report_record_id": "bpe-field-v10000-three-seed-pooled-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-fixed-26-seed137",
      "model": "Mamba3 fixed-26 seed 137",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.19708715666547383,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-fixed-26-seed137.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.30010911402708185,
        "ask_cancellation_ticks": 0.2994513204361498,
        "ask_volume": 0.04379543948724421,
        "ask_volume_touch": 0.045693795124635735,
        "bid_cancellation_depth": 0.3214880235940786,
        "bid_cancellation_ticks": 0.32035480979920516,
        "bid_volume": 0.0380339603914074,
        "bid_volume_touch": 0.046010101328391596,
        "limit_ask_order_depth": 0.20765032612726828,
        "limit_ask_order_ticks": 0.20754722982029303,
        "limit_bid_order_depth": 0.3242988878881796,
        "limit_bid_order_ticks": 0.32405219580354716,
        "log_inter_arrival_time": 0.03675109961206045,
        "log_time_to_cancel": 0.2964298823316428,
        "ofi": 0.25189045201332144,
        "ofi_down": 0.21149167798856808,
        "ofi_stay": 0.25341292604505194,
        "ofi_up": 0.19674448389753924,
        "orderbook_imbalance": 0.10716843331689224,
        "spread": 0.26066748091517544,
        "vol_per_min": 0.045788650027216105
      },
      "evaluation_key": "b7b01c18b6801817332f2bd557fa394dbed92a5416e777f318b61b55fed14c73",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.19708715666547383,
      "g20_classification": {
        "id": "bpe-fixed-26-seed137",
        "class": "G20",
        "model_evaluation": "Mamba3 fixed-26 seed 137 | step 30000; pooled January validation",
        "wws21_mean": 0.19708715666547383,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/2",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 656,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.19708715666547383,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-fixed-26-seed137.json"
            }
          ],
          "evaluation_key": "b7b01c18b6801817332f2bd557fa394dbed92a5416e777f318b61b55fed14c73",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/fixed-26-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/fixed-26-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "3e5f0ee59dfd4df652300033566c2e87c578ee17dc4008f6e94f8e4a98ff90b7",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19708715666547383,
        "feature_points": {
          "ask_cancellation_depth": 0.30010911402708185,
          "ask_cancellation_ticks": 0.2994513204361498,
          "ask_volume": 0.04379543948724421,
          "ask_volume_touch": 0.045693795124635735,
          "bid_cancellation_depth": 0.3214880235940786,
          "bid_cancellation_ticks": 0.32035480979920516,
          "bid_volume": 0.0380339603914074,
          "bid_volume_touch": 0.046010101328391596,
          "limit_ask_order_depth": 0.20765032612726828,
          "limit_ask_order_ticks": 0.20754722982029303,
          "limit_bid_order_depth": 0.3242988878881796,
          "limit_bid_order_ticks": 0.32405219580354716,
          "log_inter_arrival_time": 0.03675109961206045,
          "log_time_to_cancel": 0.2964298823316428,
          "ofi": 0.25189045201332144,
          "ofi_down": 0.21149167798856808,
          "ofi_stay": 0.25341292604505194,
          "ofi_up": 0.19674448389753924,
          "orderbook_imbalance": 0.10716843331689224,
          "spread": 0.26066748091517544,
          "vol_per_min": 0.045788650027216105
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/2",
            "source_value": 0.19708715666547383,
            "report_record_id": "bpe-fixed-26-seed137",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-fixed-26-seed137].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-fixed-26-seed137-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "fixed-26 seed 137 mean daily WS-21",
      "mean_ws21": 0.24857738803011498,
      "reported_value": 0.24857738803011498,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 137,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/fixed-26-seed137/scores/pooled/lobbench_summary.json",
      "upstream_sha256": "b07963880685b30b097803c68599fee9677d3d154eba535944d9e503c0cae87c",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.24857738803011498,
      "g20_classification": {
        "id": "bpe-fixed-26-seed137-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "fixed-26 seed 137 mean daily WS-21",
        "wws21_mean": 0.24857738803011498,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 657,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/fixed-26-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/fixed-26-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "3e5f0ee59dfd4df652300033566c2e87c578ee17dc4008f6e94f8e4a98ff90b7",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.24857738803011498,
            "report_record_id": "bpe-fixed-26-seed137-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-fixed-26-seed42",
      "model": "Mamba3 fixed-26 seed 42",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.20339953653402024,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-fixed-26-seed42.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.29558249420990734,
        "ask_cancellation_ticks": 0.29505620249558207,
        "ask_volume": 0.05061990134292658,
        "ask_volume_touch": 0.05137250324850213,
        "bid_cancellation_depth": 0.3094290773787388,
        "bid_cancellation_ticks": 0.30853795601236905,
        "bid_volume": 0.04852808342740549,
        "bid_volume_touch": 0.040775287211398156,
        "limit_ask_order_depth": 0.2141774179252639,
        "limit_ask_order_ticks": 0.21415948448957609,
        "limit_bid_order_depth": 0.31700935397735214,
        "limit_bid_order_ticks": 0.31678034997707505,
        "log_inter_arrival_time": 0.0411534563939409,
        "log_time_to_cancel": 0.36946386343729953,
        "ofi": 0.277122927425127,
        "ofi_down": 0.22310175956854167,
        "ofi_stay": 0.2786965195691154,
        "ofi_up": 0.21875111310738815,
        "orderbook_imbalance": 0.09705915642789634,
        "spread": 0.2707940510876077,
        "vol_per_min": 0.03321930850141176
      },
      "evaluation_key": "9dbe2229518689a987f91408957abc19ea67d369ff9d24299e2ac92a6ea493cf",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.20339953653402024,
      "g20_classification": {
        "id": "bpe-fixed-26-seed42",
        "class": "G20",
        "model_evaluation": "Mamba3 fixed-26 seed 42 | step 30000; pooled January validation",
        "wws21_mean": 0.20339953653402024,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/0",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 658,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.20339953653402024,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-fixed-26-seed42.json"
            }
          ],
          "evaluation_key": "9dbe2229518689a987f91408957abc19ea67d369ff9d24299e2ac92a6ea493cf",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "No checkpoint path found in accessible metadata. The canonical aggregate points to this seed-42 source summary, but both that summary and its artifact directory are permission denied. Step 30000 is recorded; a path has not been inferred from the model name.",
        "completeness": "unresolved",
        "expected_component_count": 1,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seed": 42,
        "source_metadata_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
        "source_metadata_availability": "permission_denied"
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.20339953653402024,
        "feature_points": {
          "ask_cancellation_depth": 0.29558249420990734,
          "ask_cancellation_ticks": 0.29505620249558207,
          "ask_volume": 0.05061990134292658,
          "ask_volume_touch": 0.05137250324850213,
          "bid_cancellation_depth": 0.3094290773787388,
          "bid_cancellation_ticks": 0.30853795601236905,
          "bid_volume": 0.04852808342740549,
          "bid_volume_touch": 0.040775287211398156,
          "limit_ask_order_depth": 0.2141774179252639,
          "limit_ask_order_ticks": 0.21415948448957609,
          "limit_bid_order_depth": 0.31700935397735214,
          "limit_bid_order_ticks": 0.31678034997707505,
          "log_inter_arrival_time": 0.0411534563939409,
          "log_time_to_cancel": 0.36946386343729953,
          "ofi": 0.277122927425127,
          "ofi_down": 0.22310175956854167,
          "ofi_stay": 0.2786965195691154,
          "ofi_up": 0.21875111310738815,
          "orderbook_imbalance": 0.09705915642789634,
          "spread": 0.2707940510876077,
          "vol_per_min": 0.03321930850141176
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/0",
            "source_value": 0.20339953653402024,
            "report_record_id": "bpe-fixed-26-seed42",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-fixed-26-seed42].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-fixed-26-seed42-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "fixed-26 seed 42 mean daily WS-21",
      "mean_ws21": 0.25161432088616886,
      "reported_value": 0.25161432088616886,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 42,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
      "upstream_sha256": "aa204921d690be4037279d0c280c81897945b8161fae9f4b73956aff23106aac",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.25161432088616886,
      "g20_classification": {
        "id": "bpe-fixed-26-seed42-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "fixed-26 seed 42 mean daily WS-21",
        "wws21_mean": 0.25161432088616886,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 659,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "No checkpoint path found in accessible metadata. The canonical aggregate points to this seed-42 source summary, but both that summary and its artifact directory are permission denied. Step 30000 is recorded; a path has not been inferred from the model name.",
        "completeness": "unresolved",
        "expected_component_count": 1,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seed": 42,
        "source_metadata_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
        "source_metadata_availability": "permission_denied"
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.25161432088616886,
            "report_record_id": "bpe-fixed-26-seed42-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-fixed-26-seed5",
      "model": "Mamba3 fixed-26 seed 5",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.20119797105169474,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-fixed-26-seed5.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.30208271150531374,
        "ask_cancellation_ticks": 0.30130575069728066,
        "ask_volume": 0.035660224851652905,
        "ask_volume_touch": 0.03487200053577229,
        "bid_cancellation_depth": 0.30830330774797077,
        "bid_cancellation_ticks": 0.30726344239529313,
        "bid_volume": 0.03884895626570225,
        "bid_volume_touch": 0.03835070382598694,
        "limit_ask_order_depth": 0.20709396407486297,
        "limit_ask_order_ticks": 0.20691695962546192,
        "limit_bid_order_depth": 0.31820913260984895,
        "limit_bid_order_ticks": 0.3179571704522942,
        "log_inter_arrival_time": 0.042930608435058734,
        "log_time_to_cancel": 0.3926406272960134,
        "ofi": 0.2670397180363256,
        "ofi_down": 0.23977055294263383,
        "ofi_stay": 0.26762570958151066,
        "ofi_up": 0.2204748285458693,
        "orderbook_imbalance": 0.10360712046835367,
        "spread": 0.23614713793765163,
        "vol_per_min": 0.03805676425473191
      },
      "evaluation_key": "ede9731e2d98d8ea66163455740ede0115add9b50ddf4f008597cf569d9fa970",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.20119797105169474,
      "g20_classification": {
        "id": "bpe-fixed-26-seed5",
        "class": "G20",
        "model_evaluation": "Mamba3 fixed-26 seed 5 | step 30000; pooled January validation",
        "wws21_mean": 0.20119797105169474,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/1",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 660,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.20119797105169474,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-fixed-26-seed5.json"
            }
          ],
          "evaluation_key": "ede9731e2d98d8ea66163455740ede0115add9b50ddf4f008597cf569d9fa970",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/arm-specs/generation/fixed-26-seed5/arm-0.env",
            "evidence_line": 3,
            "evidence_text": "CHECKPOINT_ROOT=/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/campaign.json",
            "checkpoint_step_evidence_line": 9,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "b4ba69d8e0ca6e81c17dbdb442a551ca9f7d92cad5b31b984d2c42a56fec0019",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.20119797105169474,
        "feature_points": {
          "ask_cancellation_depth": 0.30208271150531374,
          "ask_cancellation_ticks": 0.30130575069728066,
          "ask_volume": 0.035660224851652905,
          "ask_volume_touch": 0.03487200053577229,
          "bid_cancellation_depth": 0.30830330774797077,
          "bid_cancellation_ticks": 0.30726344239529313,
          "bid_volume": 0.03884895626570225,
          "bid_volume_touch": 0.03835070382598694,
          "limit_ask_order_depth": 0.20709396407486297,
          "limit_ask_order_ticks": 0.20691695962546192,
          "limit_bid_order_depth": 0.31820913260984895,
          "limit_bid_order_ticks": 0.3179571704522942,
          "log_inter_arrival_time": 0.042930608435058734,
          "log_time_to_cancel": 0.3926406272960134,
          "ofi": 0.2670397180363256,
          "ofi_down": 0.23977055294263383,
          "ofi_stay": 0.26762570958151066,
          "ofi_up": 0.2204748285458693,
          "orderbook_imbalance": 0.10360712046835367,
          "spread": 0.23614713793765163,
          "vol_per_min": 0.03805676425473191
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/1",
            "source_value": 0.20119797105169474,
            "report_record_id": "bpe-fixed-26-seed5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-fixed-26-seed5].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-fixed-26-seed5-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "fixed-26 seed 5 mean daily WS-21",
      "mean_ws21": 0.2553750711732196,
      "reported_value": 0.2553750711732196,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 5,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/scores/fixed-26-seed5/pooled/lobbench_summary.json",
      "upstream_sha256": "2a1406003159d2f4fbd263b30b331fc2986abd78f9ee4913dc09baa989d8b508",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2553750711732196,
      "g20_classification": {
        "id": "bpe-fixed-26-seed5-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "fixed-26 seed 5 mean daily WS-21",
        "wws21_mean": 0.2553750711732196,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 661,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/arm-specs/generation/fixed-26-seed5/arm-0.env",
            "evidence_line": 3,
            "evidence_text": "CHECKPOINT_ROOT=/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/campaign.json",
            "checkpoint_step_evidence_line": 9,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "b4ba69d8e0ca6e81c17dbdb442a551ca9f7d92cad5b31b984d2c42a56fec0019",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.2553750711732196,
            "report_record_id": "bpe-fixed-26-seed5-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-fixed-26-three-seed-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "fixed-26 three-seed mean of per-seed daily WS-21 means",
      "mean_ws21": 0.2518555933631678,
      "reported_value": 0.2518555933631678,
      "metric_name": "daily_ws21_mean.mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean across three independently trained seeds of each seed mean over daily WS-21 scores",
      "protocol_scope": "GOOG January validation; seeds 42, 5, 137; 20 trading days; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "reported_sample_sd": 0.0034052581891290045,
      "notes": "exact source values in seed order [42,5,137]: [0.25161432088616886, 0.2553750711732196, 0.24857738803011498]",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2518555933631678,
      "g20_classification": {
        "id": "bpe-fixed-26-three-seed-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "fixed-26 three-seed mean of per-seed daily WS-21 means",
        "wws21_mean": 0.2518555933631678,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean across three independently trained seeds of each seed mean over daily WS-21 scores；GOOG January validation; seeds 42, 5, 137; 20 trading days; 3,136 windows/seed; generation seed 2026；exact source values in seed order [42,5,137]: [0.25161432088616886, 0.2553750711732196, 0.24857738803011498]；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 662,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/arm-specs/generation/fixed-26-seed5/arm-0.env",
            "evidence_line": 3,
            "evidence_text": "CHECKPOINT_ROOT=/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/campaign.json",
            "checkpoint_step_evidence_line": 9,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "b4ba69d8e0ca6e81c17dbdb442a551ca9f7d92cad5b31b984d2c42a56fec0019",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/fixed-26-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/fixed-26-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "3e5f0ee59dfd4df652300033566c2e87c578ee17dc4008f6e94f8e4a98ff90b7",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run"
          }
        ],
        "note": "Partial checkpoint mapping: the aggregate contains seeds 5, 42 and 137. Recorded roots and selected step 30000 are resolved for seeds 5 and 137. The seed-42 source summary and artifact directory are permission denied, so its checkpoint path remains unresolved.",
        "completeness": "partial",
        "expected_component_count": 3,
        "resolved_component_count": 2,
        "unresolved_model_seeds": [
          42
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. The aggregate uses a common C250/G250 protocol across all three components. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.2518555933631678,
            "report_record_id": "bpe-fixed-26-three-seed-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-fixed-26-three-seed-pooled-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "fixed-26 three-seed mean pooled WS-21",
      "mean_ws21": 0.20056155475039628,
      "reported_value": 0.20056155475039628,
      "metric_name": "ws21.mean",
      "metric_kind": "unconditional_ws21",
      "aggregation_level": "mean across three independently scored model seeds",
      "protocol_scope": "GOOG January validation; pooled observations; seeds 42, 5, 137; 3,136 windows/seed; 20 trading days; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "reported_sample_sd": 0.003203951338991638,
      "notes": "exact source values in seed order [42,5,137]: [0.20339953653402024, 0.20119797105169474, 0.19708715666547383]",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.20056155475039628,
      "g20_classification": {
        "id": "bpe-fixed-26-three-seed-pooled-ws21-mean",
        "class": "G20",
        "model_evaluation": "fixed-26 three-seed mean pooled WS-21",
        "wws21_mean": 0.20056155475039628,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean across three independently scored model seeds；GOOG January validation; pooled observations; seeds 42, 5, 137; 3,136 windows/seed; 20 trading days; generation seed 2026；exact source values in seed order [42,5,137]: [0.20339953653402024, 0.20119797105169474, 0.19708715666547383]；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 663,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/arm-specs/generation/fixed-26-seed5/arm-0.env",
            "evidence_line": 3,
            "evidence_text": "CHECKPOINT_ROOT=/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_seed5_completed_v1/campaign.json",
            "checkpoint_step_evidence_line": 9,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "b4ba69d8e0ca6e81c17dbdb442a551ca9f7d92cad5b31b984d2c42a56fec0019",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/fixed-26-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/fixed-26-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "3e5f0ee59dfd4df652300033566c2e87c578ee17dc4008f6e94f8e4a98ff90b7",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_fixed_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0002/run"
          }
        ],
        "note": "Partial checkpoint mapping: the aggregate contains seeds 5, 42 and 137. Recorded roots and selected step 30000 are resolved for seeds 5 and 137. The seed-42 source summary and artifact directory are permission denied, so its checkpoint path remains unresolved.",
        "completeness": "partial",
        "expected_component_count": 3,
        "resolved_component_count": 2,
        "unresolved_model_seeds": [
          42
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. The aggregate uses a common C250/G250 protocol across all three components.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.20056155475039628,
            "report_record_id": "bpe-fixed-26-three-seed-pooled-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v15000-seed137",
      "model": "Mamba3 message-v15000 seed 137",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.2043290711264578,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v15000-seed137.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.22792718847001536,
        "ask_cancellation_ticks": 0.22764242612232674,
        "ask_volume": 0.049772397862594596,
        "ask_volume_touch": 0.05652698661148815,
        "bid_cancellation_depth": 0.316441124480504,
        "bid_cancellation_ticks": 0.31655797875749236,
        "bid_volume": 0.04811807859581543,
        "bid_volume_touch": 0.03794714417614131,
        "limit_ask_order_depth": 0.246380673136873,
        "limit_ask_order_ticks": 0.24655829578783758,
        "limit_bid_order_depth": 0.3336665371500511,
        "limit_bid_order_ticks": 0.3327329763657688,
        "log_inter_arrival_time": 0.10533610375254059,
        "log_time_to_cancel": 0.478373244697107,
        "ofi": 0.26858286073408877,
        "ofi_down": 0.17035075871973215,
        "ofi_stay": 0.2712165243960737,
        "ofi_up": 0.14246081520112575,
        "orderbook_imbalance": 0.09856949923348998,
        "spread": 0.2737286539606741,
        "vol_per_min": 0.04202022544387337
      },
      "evaluation_key": "49c419da5d1f20ba521b7dd6b400d63918c404924ecf57c4c2067e749d09183f",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2043290711264578,
      "g20_classification": {
        "id": "bpe-message-v15000-seed137",
        "class": "G20",
        "model_evaluation": "Mamba3 message-v15000 seed 137 | step 30000; pooled January validation",
        "wws21_mean": 0.2043290711264578,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/11",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 664,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.2043290711264578,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v15000-seed137.json"
            }
          ],
          "evaluation_key": "49c419da5d1f20ba521b7dd6b400d63918c404924ecf57c4c2067e749d09183f",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "d733f22c9b98f7aaf938a96c701a5f208c38bb5de49a6c86969b334ee9aa9f1f",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.2043290711264578,
        "feature_points": {
          "ask_cancellation_depth": 0.22792718847001536,
          "ask_cancellation_ticks": 0.22764242612232674,
          "ask_volume": 0.049772397862594596,
          "ask_volume_touch": 0.05652698661148815,
          "bid_cancellation_depth": 0.316441124480504,
          "bid_cancellation_ticks": 0.31655797875749236,
          "bid_volume": 0.04811807859581543,
          "bid_volume_touch": 0.03794714417614131,
          "limit_ask_order_depth": 0.246380673136873,
          "limit_ask_order_ticks": 0.24655829578783758,
          "limit_bid_order_depth": 0.3336665371500511,
          "limit_bid_order_ticks": 0.3327329763657688,
          "log_inter_arrival_time": 0.10533610375254059,
          "log_time_to_cancel": 0.478373244697107,
          "ofi": 0.26858286073408877,
          "ofi_down": 0.17035075871973215,
          "ofi_stay": 0.2712165243960737,
          "ofi_up": 0.14246081520112575,
          "orderbook_imbalance": 0.09856949923348998,
          "spread": 0.2737286539606741,
          "vol_per_min": 0.04202022544387337
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/11",
            "source_value": 0.2043290711264578,
            "report_record_id": "bpe-message-v15000-seed137",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-message-v15000-seed137].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v15000-seed137-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v15000 seed 137 mean daily WS-21",
      "mean_ws21": 0.2631824093063037,
      "reported_value": 0.2631824093063037,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 137,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed137/scores/pooled/lobbench_summary.json",
      "upstream_sha256": "d0c867e54bba0306a32d5d8ac470d2165da63847ce79fa94394a9b5d6bf81db2",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2631824093063037,
      "g20_classification": {
        "id": "bpe-message-v15000-seed137-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v15000 seed 137 mean daily WS-21",
        "wws21_mean": 0.2631824093063037,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 665,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "d733f22c9b98f7aaf938a96c701a5f208c38bb5de49a6c86969b334ee9aa9f1f",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.2631824093063037,
            "report_record_id": "bpe-message-v15000-seed137-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v15000-seed42",
      "model": "Mamba3 message-v15000 seed 42",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.17604628741347506,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v15000-seed42.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.25988857243322805,
        "ask_cancellation_ticks": 0.2595309313522621,
        "ask_volume": 0.01122185388795447,
        "ask_volume_touch": 0.011189581690935483,
        "bid_cancellation_depth": 0.3064477539318645,
        "bid_cancellation_ticks": 0.3065013578145602,
        "bid_volume": 0.0467491461034825,
        "bid_volume_touch": 0.05077198683871022,
        "limit_ask_order_depth": 0.22115489239643915,
        "limit_ask_order_ticks": 0.22125181862990714,
        "limit_bid_order_depth": 0.3274139442599233,
        "limit_bid_order_ticks": 0.32641406186952715,
        "log_inter_arrival_time": 0.07302092180945716,
        "log_time_to_cancel": 0.43626690607046914,
        "ofi": 0.014481837221898271,
        "ofi_down": 0.16600908438445316,
        "ofi_stay": 0.014508050701116899,
        "ofi_up": 0.20171044980160457,
        "orderbook_imbalance": 0.11055186906473222,
        "spread": 0.28286201630426994,
        "vol_per_min": 0.04902499911618038
      },
      "evaluation_key": "65f40d2ce42f335960e4489092993336ed9a25dcd2d4fa592b0851ed2e963eb2",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.17604628741347506,
      "g20_classification": {
        "id": "bpe-message-v15000-seed42",
        "class": "G20",
        "model_evaluation": "Mamba3 message-v15000 seed 42 | step 30000; pooled January validation",
        "wws21_mean": 0.17604628741347506,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/9",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 666,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.17604628741347503,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v15000-seed42.json"
            }
          ],
          "evaluation_key": "65f40d2ce42f335960e4489092993336ed9a25dcd2d4fa592b0851ed2e963eb2",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "No checkpoint path found in accessible metadata. The canonical aggregate points to this seed-42 source summary, but both that summary and its artifact directory are permission denied. Step 30000 is recorded; a path has not been inferred from the model name.",
        "completeness": "unresolved",
        "expected_component_count": 1,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seed": 42,
        "source_metadata_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
        "source_metadata_availability": "permission_denied"
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.17604628741347503,
        "feature_points": {
          "ask_cancellation_depth": 0.25988857243322805,
          "ask_cancellation_ticks": 0.2595309313522621,
          "ask_volume": 0.01122185388795447,
          "ask_volume_touch": 0.011189581690935483,
          "bid_cancellation_depth": 0.3064477539318645,
          "bid_cancellation_ticks": 0.3065013578145602,
          "bid_volume": 0.0467491461034825,
          "bid_volume_touch": 0.05077198683871022,
          "limit_ask_order_depth": 0.22115489239643915,
          "limit_ask_order_ticks": 0.22125181862990714,
          "limit_bid_order_depth": 0.3274139442599233,
          "limit_bid_order_ticks": 0.32641406186952715,
          "log_inter_arrival_time": 0.07302092180945716,
          "log_time_to_cancel": 0.43626690607046914,
          "ofi": 0.014481837221898271,
          "ofi_down": 0.16600908438445316,
          "ofi_stay": 0.014508050701116899,
          "ofi_up": 0.20171044980160457,
          "orderbook_imbalance": 0.11055186906473222,
          "spread": 0.28286201630426994,
          "vol_per_min": 0.04902499911618038
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/9",
            "source_value": 0.17604628741347506,
            "report_record_id": "bpe-message-v15000-seed42",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-message-v15000-seed42].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v15000-seed42-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v15000 seed 42 mean daily WS-21",
      "mean_ws21": 0.25745307927208777,
      "reported_value": 0.25745307927208777,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 42,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
      "upstream_sha256": "aa204921d690be4037279d0c280c81897945b8161fae9f4b73956aff23106aac",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.25745307927208777,
      "g20_classification": {
        "id": "bpe-message-v15000-seed42-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v15000 seed 42 mean daily WS-21",
        "wws21_mean": 0.25745307927208777,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 667,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "No checkpoint path found in accessible metadata. The canonical aggregate points to this seed-42 source summary, but both that summary and its artifact directory are permission denied. Step 30000 is recorded; a path has not been inferred from the model name.",
        "completeness": "unresolved",
        "expected_component_count": 1,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seed": 42,
        "source_metadata_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
        "source_metadata_availability": "permission_denied"
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.25745307927208777,
            "report_record_id": "bpe-message-v15000-seed42-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v15000-seed5",
      "model": "Mamba3 message-v15000 seed 5",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.22294189658687966,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v15000-seed5.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.2834590021062557,
        "ask_cancellation_ticks": 0.2832783610064551,
        "ask_volume": 0.05564282718930476,
        "ask_volume_touch": 0.055378232461395976,
        "bid_cancellation_depth": 0.3381955112438897,
        "bid_cancellation_ticks": 0.338487398670216,
        "bid_volume": 0.06505613003760963,
        "bid_volume_touch": 0.04013358138988575,
        "limit_ask_order_depth": 0.2475508415804495,
        "limit_ask_order_ticks": 0.24771091355249536,
        "limit_bid_order_depth": 0.3456015825023335,
        "limit_bid_order_ticks": 0.34474921721540064,
        "log_inter_arrival_time": 0.115445494031798,
        "log_time_to_cancel": 0.4640341532688914,
        "ofi": 0.3117965628154336,
        "ofi_down": 0.19556776654125674,
        "ofi_stay": 0.3135246790105112,
        "ofi_up": 0.22611446978235247,
        "orderbook_imbalance": 0.10322995179370469,
        "spread": 0.2643602101089362,
        "vol_per_min": 0.04246294201589709
      },
      "evaluation_key": "1f566dbf8f1d2df7f01284abd0b70664f8ce2f4b3c3e2a46dd356cf7245739fd",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.22294189658687966,
      "g20_classification": {
        "id": "bpe-message-v15000-seed5",
        "class": "G20",
        "model_evaluation": "Mamba3 message-v15000 seed 5 | step 30000; pooled January validation",
        "wws21_mean": 0.22294189658687966,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/10",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 668,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.22294189658687966,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v15000-seed5.json"
            }
          ],
          "evaluation_key": "1f566dbf8f1d2df7f01284abd0b70664f8ce2f4b3c3e2a46dd356cf7245739fd",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed5/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed5/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "b21445e9032534493ae48270b6c3b883444044ab6acccff03ee7ceda5a79866c",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.22294189658687966,
        "feature_points": {
          "ask_cancellation_depth": 0.2834590021062557,
          "ask_cancellation_ticks": 0.2832783610064551,
          "ask_volume": 0.05564282718930476,
          "ask_volume_touch": 0.055378232461395976,
          "bid_cancellation_depth": 0.3381955112438897,
          "bid_cancellation_ticks": 0.338487398670216,
          "bid_volume": 0.06505613003760963,
          "bid_volume_touch": 0.04013358138988575,
          "limit_ask_order_depth": 0.2475508415804495,
          "limit_ask_order_ticks": 0.24771091355249536,
          "limit_bid_order_depth": 0.3456015825023335,
          "limit_bid_order_ticks": 0.34474921721540064,
          "log_inter_arrival_time": 0.115445494031798,
          "log_time_to_cancel": 0.4640341532688914,
          "ofi": 0.3117965628154336,
          "ofi_down": 0.19556776654125674,
          "ofi_stay": 0.3135246790105112,
          "ofi_up": 0.22611446978235247,
          "orderbook_imbalance": 0.10322995179370469,
          "spread": 0.2643602101089362,
          "vol_per_min": 0.04246294201589709
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/10",
            "source_value": 0.22294189658687966,
            "report_record_id": "bpe-message-v15000-seed5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-message-v15000-seed5].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v15000-seed5-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v15000 seed 5 mean daily WS-21",
      "mean_ws21": 0.27428051545272336,
      "reported_value": 0.27428051545272336,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 5,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed5/scores/pooled/lobbench_summary.json",
      "upstream_sha256": "7230c086224e8ca4ebd88306d921db4a93aee217e3e3bb30c40d8e69a914d3b8",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.27428051545272336,
      "g20_classification": {
        "id": "bpe-message-v15000-seed5-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v15000 seed 5 mean daily WS-21",
        "wws21_mean": 0.27428051545272336,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 669,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed5/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed5/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "b21445e9032534493ae48270b6c3b883444044ab6acccff03ee7ceda5a79866c",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.27428051545272336,
            "report_record_id": "bpe-message-v15000-seed5-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v15000-three-seed-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v15000 three-seed mean of per-seed daily WS-21 means",
      "mean_ws21": 0.26497200134370497,
      "reported_value": 0.26497200134370497,
      "metric_name": "daily_ws21_mean.mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean across three independently trained seeds of each seed mean over daily WS-21 scores",
      "protocol_scope": "GOOG January validation; seeds 42, 5, 137; 20 trading days; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "reported_sample_sd": 0.008555269244657842,
      "notes": "exact source values in seed order [42,5,137]: [0.25745307927208777, 0.27428051545272336, 0.2631824093063037]",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.26497200134370497,
      "g20_classification": {
        "id": "bpe-message-v15000-three-seed-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v15000 three-seed mean of per-seed daily WS-21 means",
        "wws21_mean": 0.26497200134370497,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean across three independently trained seeds of each seed mean over daily WS-21 scores；GOOG January validation; seeds 42, 5, 137; 20 trading days; 3,136 windows/seed; generation seed 2026；exact source values in seed order [42,5,137]: [0.25745307927208777, 0.27428051545272336, 0.2631824093063037]；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 670,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed5/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed5/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "b21445e9032534493ae48270b6c3b883444044ab6acccff03ee7ceda5a79866c",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "d733f22c9b98f7aaf938a96c701a5f208c38bb5de49a6c86969b334ee9aa9f1f",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run"
          }
        ],
        "note": "Partial checkpoint mapping: the aggregate contains seeds 5, 42 and 137. Recorded roots and selected step 30000 are resolved for seeds 5 and 137. The seed-42 source summary and artifact directory are permission denied, so its checkpoint path remains unresolved.",
        "completeness": "partial",
        "expected_component_count": 3,
        "resolved_component_count": 2,
        "unresolved_model_seeds": [
          42
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. The aggregate uses a common C250/G250 protocol across all three components. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.26497200134370497,
            "report_record_id": "bpe-message-v15000-three-seed-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v15000-three-seed-pooled-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v15000 three-seed mean pooled WS-21",
      "mean_ws21": 0.2011057517089375,
      "reported_value": 0.2011057517089375,
      "metric_name": "ws21.mean",
      "metric_kind": "unconditional_ws21",
      "aggregation_level": "mean across three independently scored model seeds",
      "protocol_scope": "GOOG January validation; pooled observations; seeds 42, 5, 137; 3,136 windows/seed; 20 trading days; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "reported_sample_sd": 0.023613383514157826,
      "notes": "exact source values in seed order [42,5,137]: [0.17604628741347503, 0.22294189658687966, 0.2043290711264578]",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2011057517089375,
      "g20_classification": {
        "id": "bpe-message-v15000-three-seed-pooled-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v15000 three-seed mean pooled WS-21",
        "wws21_mean": 0.2011057517089375,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean across three independently scored model seeds；GOOG January validation; pooled observations; seeds 42, 5, 137; 3,136 windows/seed; 20 trading days; generation seed 2026；exact source values in seed order [42,5,137]: [0.17604628741347503, 0.22294189658687966, 0.2043290711264578]；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 671,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed5/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed5/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "b21445e9032534493ae48270b6c3b883444044ab6acccff03ee7ceda5a79866c",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0002/run"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v15000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "d733f22c9b98f7aaf938a96c701a5f208c38bb5de49a6c86969b334ee9aa9f1f",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v15000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run"
          }
        ],
        "note": "Partial checkpoint mapping: the aggregate contains seeds 5, 42 and 137. Recorded roots and selected step 30000 are resolved for seeds 5 and 137. The seed-42 source summary and artifact directory are permission denied, so its checkpoint path remains unresolved.",
        "completeness": "partial",
        "expected_component_count": 3,
        "resolved_component_count": 2,
        "unresolved_model_seeds": [
          42
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. The aggregate uses a common C250/G250 protocol across all three components.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.2011057517089375,
            "report_record_id": "bpe-message-v15000-three-seed-pooled-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v5000-seed137",
      "model": "Mamba3 message-v5000 seed 137",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.19192768239804625,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v5000-seed137.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.21586330563546205,
        "ask_cancellation_ticks": 0.21567620461905618,
        "ask_volume": 0.0669347962235085,
        "ask_volume_touch": 0.0509131823630544,
        "bid_cancellation_depth": 0.2912013048649491,
        "bid_cancellation_ticks": 0.29105138396818964,
        "bid_volume": 0.05368848019970948,
        "bid_volume_touch": 0.03747509704707104,
        "limit_ask_order_depth": 0.15997422337961936,
        "limit_ask_order_ticks": 0.16007703443350552,
        "limit_bid_order_depth": 0.31275316364840633,
        "limit_bid_order_ticks": 0.31176227860460415,
        "log_inter_arrival_time": 0.08914863802329814,
        "log_time_to_cancel": 0.41707467034012136,
        "ofi": 0.27930918316908726,
        "ofi_down": 0.1939316438643737,
        "ofi_stay": 0.28076190598337436,
        "ofi_up": 0.20874948570566443,
        "orderbook_imbalance": 0.09449351450557336,
        "spread": 0.2565817312608426,
        "vol_per_min": 0.043060102519500584
      },
      "evaluation_key": "373a5c62cfc80826a97e5fad3b99038a7b9b20633347d574c83be37e3d6ac5ae",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.19192768239804625,
      "g20_classification": {
        "id": "bpe-message-v5000-seed137",
        "class": "G20",
        "model_evaluation": "Mamba3 message-v5000 seed 137 | step 30000; pooled January validation",
        "wws21_mean": 0.19192768239804625,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/8",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 672,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.19192768239804625,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v5000-seed137.json"
            }
          ],
          "evaluation_key": "373a5c62cfc80826a97e5fad3b99038a7b9b20633347d574c83be37e3d6ac5ae",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "491adf3a730c1626e88e5cecbcb742a4472cfda07883132eeab9a464a0fa7bc2",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19192768239804625,
        "feature_points": {
          "ask_cancellation_depth": 0.21586330563546205,
          "ask_cancellation_ticks": 0.21567620461905618,
          "ask_volume": 0.0669347962235085,
          "ask_volume_touch": 0.0509131823630544,
          "bid_cancellation_depth": 0.2912013048649491,
          "bid_cancellation_ticks": 0.29105138396818964,
          "bid_volume": 0.05368848019970948,
          "bid_volume_touch": 0.03747509704707104,
          "limit_ask_order_depth": 0.15997422337961936,
          "limit_ask_order_ticks": 0.16007703443350552,
          "limit_bid_order_depth": 0.31275316364840633,
          "limit_bid_order_ticks": 0.31176227860460415,
          "log_inter_arrival_time": 0.08914863802329814,
          "log_time_to_cancel": 0.41707467034012136,
          "ofi": 0.27930918316908726,
          "ofi_down": 0.1939316438643737,
          "ofi_stay": 0.28076190598337436,
          "ofi_up": 0.20874948570566443,
          "orderbook_imbalance": 0.09449351450557336,
          "spread": 0.2565817312608426,
          "vol_per_min": 0.043060102519500584
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/8",
            "source_value": 0.19192768239804625,
            "report_record_id": "bpe-message-v5000-seed137",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-message-v5000-seed137].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v5000-seed137-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v5000 seed 137 mean daily WS-21",
      "mean_ws21": 0.26463312027492564,
      "reported_value": 0.26463312027492564,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 137,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed137/scores/pooled/lobbench_summary.json",
      "upstream_sha256": "a6047a1b1b96adb66bf0cd7438ce71c28650432eb7f701bff91edbe65d95bb3c",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.26463312027492564,
      "g20_classification": {
        "id": "bpe-message-v5000-seed137-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v5000 seed 137 mean daily WS-21",
        "wws21_mean": 0.26463312027492564,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 673,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "491adf3a730c1626e88e5cecbcb742a4472cfda07883132eeab9a464a0fa7bc2",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.26463312027492564,
            "report_record_id": "bpe-message-v5000-seed137-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v5000-seed42",
      "model": "Mamba3 message-v5000 seed 42",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.1840771482514255,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v5000-seed42.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.10951884192579575,
        "ask_cancellation_ticks": 0.10929051572910536,
        "ask_volume": 0.0514462962853693,
        "ask_volume_touch": 0.055208900612585196,
        "bid_cancellation_depth": 0.3137203445365988,
        "bid_cancellation_ticks": 0.3138201212475552,
        "bid_volume": 0.06382166858221566,
        "bid_volume_touch": 0.04258913652814453,
        "limit_ask_order_depth": 0.20912710523097272,
        "limit_ask_order_ticks": 0.20921254281757404,
        "limit_bid_order_depth": 0.3285985401460026,
        "limit_bid_order_ticks": 0.327714885292569,
        "log_inter_arrival_time": 0.06732658931137286,
        "log_time_to_cancel": 0.4066886847581437,
        "ofi": 0.23983429636128384,
        "ofi_down": 0.18445381993872878,
        "ofi_stay": 0.24187885919397425,
        "ofi_up": 0.163699864296614,
        "orderbook_imbalance": 0.10378481113464574,
        "spread": 0.2792011533220524,
        "vol_per_min": 0.044683136028631235
      },
      "evaluation_key": "0940e47c4a17e2c39e648eff3029efb0ce5aeb6254121f6214139bb223e19f59",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.1840771482514255,
      "g20_classification": {
        "id": "bpe-message-v5000-seed42",
        "class": "G20",
        "model_evaluation": "Mamba3 message-v5000 seed 42 | step 30000; pooled January validation",
        "wws21_mean": 0.1840771482514255,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/6",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 674,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.1840771482514255,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v5000-seed42.json"
            }
          ],
          "evaluation_key": "0940e47c4a17e2c39e648eff3029efb0ce5aeb6254121f6214139bb223e19f59",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "No checkpoint path found in accessible metadata. The canonical aggregate points to this seed-42 source summary, but both that summary and its artifact directory are permission denied. Step 30000 is recorded; a path has not been inferred from the model name.",
        "completeness": "unresolved",
        "expected_component_count": 1,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seed": 42,
        "source_metadata_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
        "source_metadata_availability": "permission_denied"
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1840771482514255,
        "feature_points": {
          "ask_cancellation_depth": 0.10951884192579575,
          "ask_cancellation_ticks": 0.10929051572910536,
          "ask_volume": 0.0514462962853693,
          "ask_volume_touch": 0.055208900612585196,
          "bid_cancellation_depth": 0.3137203445365988,
          "bid_cancellation_ticks": 0.3138201212475552,
          "bid_volume": 0.06382166858221566,
          "bid_volume_touch": 0.04258913652814453,
          "limit_ask_order_depth": 0.20912710523097272,
          "limit_ask_order_ticks": 0.20921254281757404,
          "limit_bid_order_depth": 0.3285985401460026,
          "limit_bid_order_ticks": 0.327714885292569,
          "log_inter_arrival_time": 0.06732658931137286,
          "log_time_to_cancel": 0.4066886847581437,
          "ofi": 0.23983429636128384,
          "ofi_down": 0.18445381993872878,
          "ofi_stay": 0.24187885919397425,
          "ofi_up": 0.163699864296614,
          "orderbook_imbalance": 0.10378481113464574,
          "spread": 0.2792011533220524,
          "vol_per_min": 0.044683136028631235
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/6",
            "source_value": 0.1840771482514255,
            "report_record_id": "bpe-message-v5000-seed42",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-message-v5000-seed42].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v5000-seed42-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v5000 seed 42 mean daily WS-21",
      "mean_ws21": 0.2502573666221897,
      "reported_value": 0.2502573666221897,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 42,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
      "upstream_sha256": "aa204921d690be4037279d0c280c81897945b8161fae9f4b73956aff23106aac",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2502573666221897,
      "g20_classification": {
        "id": "bpe-message-v5000-seed42-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v5000 seed 42 mean daily WS-21",
        "wws21_mean": 0.2502573666221897,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 675,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "No checkpoint path found in accessible metadata. The canonical aggregate points to this seed-42 source summary, but both that summary and its artifact directory are permission denied. Step 30000 is recorded; a path has not been inferred from the model name.",
        "completeness": "unresolved",
        "expected_component_count": 1,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seed": 42,
        "source_metadata_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_85156e5_selection_full_30k_v12/report_assets/lobbench_summary.json",
        "source_metadata_availability": "permission_denied"
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.2502573666221897,
            "report_record_id": "bpe-message-v5000-seed42-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v5000-seed5",
      "model": "Mamba3 message-v5000 seed 5",
      "evaluation": "step 30000; pooled January validation",
      "protocol_id": "bpe-goog-jan26",
      "mean_ws21": 0.20640622470756798,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v5000-seed5.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.2941806487267006,
        "ask_cancellation_ticks": 0.2938742013409515,
        "ask_volume": 0.05815966480130671,
        "ask_volume_touch": 0.05186806993037618,
        "bid_cancellation_depth": 0.29899791937012066,
        "bid_cancellation_ticks": 0.2989004225434222,
        "bid_volume": 0.07195249775888982,
        "bid_volume_touch": 0.05782942369698695,
        "limit_ask_order_depth": 0.21700764793175348,
        "limit_ask_order_ticks": 0.2171449959001893,
        "limit_bid_order_depth": 0.31891136438562395,
        "limit_bid_order_ticks": 0.31789190040850734,
        "log_inter_arrival_time": 0.06475180261994286,
        "log_time_to_cancel": 0.4280776086104354,
        "ofi": 0.25771437206841374,
        "ofi_down": 0.2033484836639286,
        "ofi_stay": 0.2591605624988559,
        "ofi_up": 0.21263276129803713,
        "orderbook_imbalance": 0.11243316010452355,
        "spread": 0.2582027636135053,
        "vol_per_min": 0.04149044758645637
      },
      "evaluation_key": "e8132cb311505e7bfae897be8264a3d5f89213a929e04cf28b2a38620838de10",
      "date": "2026-08-30",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.20640622470756798,
      "g20_classification": {
        "id": "bpe-message-v5000-seed5",
        "class": "G20",
        "model_evaluation": "Mamba3 message-v5000 seed 5 | step 30000; pooled January validation",
        "wws21_mean": 0.20640622470756798,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment: mamba3-bpe-three-seed-ws-8022cc4）；不作为评估完成时间；path 类型：score_summary；protocol=bpe-goog-jan26",
        "protocol": "bpe-goog-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "records/7",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 676,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.206406224707568,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/bpe-message-v5000-seed5.json"
            }
          ],
          "evaluation_key": "e8132cb311505e7bfae897be8264a3d5f89213a929e04cf28b2a38620838de10",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed5/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed5/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "df8f8a1cab5a74c9e60fe75943922c62c63108fa9710c4c674df3db00987a95d",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.206406224707568,
        "feature_points": {
          "ask_cancellation_depth": 0.2941806487267006,
          "ask_cancellation_ticks": 0.2938742013409515,
          "ask_volume": 0.05815966480130671,
          "ask_volume_touch": 0.05186806993037618,
          "bid_cancellation_depth": 0.29899791937012066,
          "bid_cancellation_ticks": 0.2989004225434222,
          "bid_volume": 0.07195249775888982,
          "bid_volume_touch": 0.05782942369698695,
          "limit_ask_order_depth": 0.21700764793175348,
          "limit_ask_order_ticks": 0.2171449959001893,
          "limit_bid_order_depth": 0.31891136438562395,
          "limit_bid_order_ticks": 0.31789190040850734,
          "log_inter_arrival_time": 0.06475180261994286,
          "log_time_to_cancel": 0.4280776086104354,
          "ofi": 0.25771437206841374,
          "ofi_down": 0.2033484836639286,
          "ofi_stay": 0.2591605624988559,
          "ofi_up": 0.21263276129803713,
          "orderbook_imbalance": 0.11243316010452355,
          "spread": 0.2582027636135053,
          "vol_per_min": 0.04149044758645637
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": "records/7",
            "source_value": 0.20640622470756798,
            "report_record_id": "bpe-message-v5000-seed5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=bpe-message-v5000-seed5].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v5000-seed5-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v5000 seed 5 mean daily WS-21",
      "mean_ws21": 0.2586497093909141,
      "reported_value": 0.2586497093909141,
      "metric_name": "daily_ws21_mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean of daily 21-feature WS-21 scores for one independently trained model seed",
      "protocol_scope": "GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "seed": 5,
      "upstream_source": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed5/scores/pooled/lobbench_summary.json",
      "upstream_sha256": "fbd3745349498cf35afea06e95dab405b00854b6664f4b34610f65a4133f2796",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.2586497093909141,
      "g20_classification": {
        "id": "bpe-message-v5000-seed5-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v5000 seed 5 mean daily WS-21",
        "wws21_mean": 0.2586497093909141,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean of daily 21-feature WS-21 scores for one independently trained model seed；GOOG January validation; 20 trading-day scores; 3,136 windows/seed; generation seed 2026；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 677,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed5/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed5/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "df8f8a1cab5a74c9e60fe75943922c62c63108fa9710c4c674df3db00987a95d",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          }
        ],
        "note": "The evaluation arm metadata explicitly records this checkpoint root; its campaign/specification pins step 30000. The same trained checkpoint underlies the pooled and daily summaries. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.2586497093909141,
            "report_record_id": "bpe-message-v5000-seed5-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v5000-three-seed-daily-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v5000 three-seed mean of per-seed daily WS-21 means",
      "mean_ws21": 0.25784673209600983,
      "reported_value": 0.25784673209600983,
      "metric_name": "daily_ws21_mean.mean",
      "metric_kind": "unconditional_ws21_daily_mean",
      "aggregation_level": "mean across three independently trained seeds of each seed mean over daily WS-21 scores",
      "protocol_scope": "GOOG January validation; seeds 42, 5, 137; 20 trading days; 3,136 windows/seed; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "reported_sample_sd": 0.007221437022721746,
      "notes": "exact source values in seed order [42,5,137]: [0.2502573666221897, 0.2586497093909141, 0.26463312027492564]",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.25784673209600983,
      "g20_classification": {
        "id": "bpe-message-v5000-three-seed-daily-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v5000 three-seed mean of per-seed daily WS-21 means",
        "wws21_mean": 0.25784673209600983,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean across three independently trained seeds of each seed mean over daily WS-21 scores；GOOG January validation; seeds 42, 5, 137; 20 trading days; 3,136 windows/seed; generation seed 2026；exact source values in seed order [42,5,137]: [0.2502573666221897, 0.2586497093909141, 0.26463312027492564]；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 678,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed5/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed5/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "df8f8a1cab5a74c9e60fe75943922c62c63108fa9710c4c674df3db00987a95d",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "491adf3a730c1626e88e5cecbcb742a4472cfda07883132eeab9a464a0fa7bc2",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run"
          }
        ],
        "note": "Partial checkpoint mapping: the aggregate contains seeds 5, 42 and 137. Recorded roots and selected step 30000 are resolved for seeds 5 and 137. The seed-42 source summary and artifact directory are permission denied, so its checkpoint path remains unresolved.",
        "completeness": "partial",
        "expected_component_count": 3,
        "resolved_component_count": 2,
        "unresolved_model_seeds": [
          42
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. The aggregate uses a common C250/G250 protocol across all three components. Daily score aggregation does not change the per-window rollout lengths.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.25784673209600983,
            "report_record_id": "bpe-message-v5000-three-seed-daily-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "bpe-message-v5000-three-seed-pooled-ws21-mean",
      "experiment_slug": "mamba3-bpe-three-seed-ws-8022cc4",
      "label": "message-v5000 three-seed mean pooled WS-21",
      "mean_ws21": 0.19413701845234657,
      "reported_value": 0.19413701845234657,
      "metric_name": "ws21.mean",
      "metric_kind": "unconditional_ws21",
      "aggregation_level": "mean across three independently scored model seeds",
      "protocol_scope": "GOOG January validation; pooled observations; seeds 42, 5, 137; 3,136 windows/seed; 20 trading days; generation seed 2026",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "date_precision": "day",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
      "reported_sample_sd": 0.011327302776772942,
      "notes": "exact source values in seed order [42,5,137]: [0.1840771482514255, 0.206406224707568, 0.19192768239804625]",
      "date": "2026-08-30",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08-30",
      "archive_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "archive_evaluated_at": "2026-08-30",
      "source_date": "2026-08-30",
      "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.19413701845234657,
      "g20_classification": {
        "id": "bpe-message-v5000-three-seed-pooled-ws21-mean",
        "class": "G20",
        "model_evaluation": "message-v5000 three-seed mean pooled WS-21",
        "wws21_mean": 0.19413701845234657,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08-30",
        "source_date_basis": "registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)",
        "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08-30（registry metadata date_experiment (artifact was created 2026-09-11 as a rendering-only aggregation; no new evaluation)）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；mean across three independently scored model seeds；GOOG January validation; pooled observations; seeds 42, 5, 137; 3,136 windows/seed; 20 trading days; generation seed 2026；exact source values in seed order [42,5,137]: [0.1840771482514255, 0.206406224707568, 0.19192768239804625]；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 679,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 5,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed5/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed5/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "df8f8a1cab5a74c9e60fe75943922c62c63108fa9710c4c674df3db00987a95d",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed5/attempts/attempt-0001/run"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "path_kind": "directory",
            "checkpoint_step": 30000,
            "model_seed": 137,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run/30000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed137/specification.json",
            "evidence_line": 3,
            "evidence_text": "\"checkpoint_root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_pareto_replicates_4cb74e5_v1/arms/message-v5000-seed137/specification.json",
            "checkpoint_step_evidence_line": 4,
            "checkpoint_step_evidence_text": "\"checkpoint_step\": 30000,",
            "recorded_checkpoint_tree_sha256": "491adf3a730c1626e88e5cecbcb742a4472cfda07883132eeab9a464a0fa7bc2",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/campaign_pareto_seed_replicates_v1/checkpoints/pareto_seed_replication/pareto_seed_replication_message_v5000_messages500_message_once_real_parameter_matched_seed137/attempts/attempt-0001/run"
          }
        ],
        "note": "Partial checkpoint mapping: the aggregate contains seeds 5, 42 and 137. Recorded roots and selected step 30000 are resolved for seeds 5 and 137. The seed-42 source summary and artifact directory are permission denied, so its checkpoint path remains unresolved.",
        "completeness": "partial",
        "expected_component_count": 3,
        "resolved_component_count": 2,
        "unresolved_model_seeds": [
          42
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/result.md",
            "line": 11,
            "text": "Four original tokenizer arms: Fixed 26, Field 10k, Message 5k and Message 15k. Model seeds 42, 5 and 137. GOOG 2022–2025 training, approximately 79–80M parameters, 30k steps. Evaluation: January 2026 GOOG, 20 days, 3,136 windows per model, condition 250 and generate 250 messages. January was used for validation/selection."
          }
        ],
        "note": "The canonical evaluation report explicitly specifies 250 conditioning and 250 generated LOB messages for all four tokenizer arms and model seeds 42, 5 and 137. These are evaluation event counts; tokenizer vocabulary, variable tokenized length and training sequence length are separate quantities. The aggregate uses a common C250/G250 protocol across all three components.",
        "evidence_scope": "Common evaluation protocol explicitly covers all tokenizer arms and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/tokenization/mamba3-bpe-three-seed-ws-8022cc4/artifacts/mamba3-bpe-three-seed-results.json",
            "selector": null,
            "source_value": 0.19413701845234657,
            "report_record_id": "bpe-message-v5000-three-seed-pooled-ws21-mean",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37512c45-68fd-814a-accb-f4889b973e71",
      "source_row_id": "37512c45-68fd-814a-accb-f4889b973e71",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37512c4568fd814aaccbf4889b973e71",
      "run": "smoke-scaling-2p63M-s42-step10610",
      "stock": null,
      "model_size": "2p63M",
      "run_family": "scaling-law",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.271284,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.271284",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.271725,
      "reported_ks_context": 0.168051,
      "n_batches": 2,
      "inventory": {
        "family": "misc/other",
        "model_size": "2p63M",
        "seed": "s42",
        "step": 10610,
        "stocks": "GOOG",
        "score_families": [
          "uncond"
        ],
        "generated_sequences": 2048,
        "run_directories": 1,
        "extended": "__NO__",
        "reported_uncond_l1_context": 0.2717,
        "source_reference": "https://app.notion.com/37412c4568fd811187a1ea9972708649"
      },
      "recorded_at": "2026-05-21T00:00:00Z",
      "date_basis": "notion_run_date",
      "date_precision": "day",
      "date_source_field": "date:Run date:start",
      "date_source_reference": "https://app.notion.com/37512c4568fd814aaccbf4889b973e71",
      "date_join_status": "not_needed",
      "date_is_evaluation_completion": false,
      "date": "2026-05-21T00:00:00Z",
      "collection_source": "notion",
      "archive_source_date": "2026-05-21T00:00:00Z",
      "archive_date_basis": "notion_run_date",
      "source_date": "2026-05-21T00:00:00Z",
      "source_date_basis": "notion_run_date",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.271284,
      "g20_classification": {
        "id": "canonical-scores:37512c45-68fd-814a-accb-f4889b973e71",
        "class": "G20",
        "model_evaluation": "smoke-scaling-2p63M-s42-step10610",
        "wws21_mean": 0.271284,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-05-21T00:00:00Z",
        "source_date_basis": "notion_run_date",
        "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/inference_results/smoke-scaling-2p63M-s42-step10610_GOOG_4686590/data_real",
        "input_path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/inference_results/smoke-scaling-2p63M-s42-step10610_GOOG_4686590/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-05-21T00:00:00Z（notion_run_date）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37512c4568fd814aaccbf4889b973e71",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 680,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4499068_9itlcx0n_4499068/10610",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/logs/integrated_smoke-scaling-2p63M-s42-step10610_GOOG_4686590.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4499068_9itlcx0n_4499068 (step 10610)",
            "checkpoint_step": 10610,
            "logged_path": "/lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4499068_9itlcx0n_4499068",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/logs/integrated_smoke-scaling-2p63M-s42-step10610_GOOG_4686590.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/public/s5e/quant_team/quant/miniforge3/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4499068_9itlcx0n_4499068\"         --save_dir \"/lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/inference_results/smoke-scaling-2p63M-s42-step10610_GOOG_4686590\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 32         --checkpoint_step 10610         --test_split 1.0 --save_format npy"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 19,
                "evidence_text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
              }
            ],
            "logged_availability": "exists",
            "availability_checked_at": "2026-09-12T06:10:01.120845+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4499068_9itlcx0n_4499068",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4499068_9itlcx0n_4499068"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/logs/integrated_smoke-scaling-2p63M-s42-step10610_GOOG_4686590.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/public/s5e/quant_team/quant/miniforge3/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4499068_9itlcx0n_4499068\"         --save_dir \"/lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/inference_results/smoke-scaling-2p63M-s42-step10610_GOOG_4686590\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 32         --checkpoint_step 10610         --test_split 1.0 --save_format npy"
          },
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/logs/integrated_smoke-scaling-2p63M-s42-step10610_GOOG_4686590.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/logs/integrated_smoke-scaling-2p63M-s42-step10610_GOOG_4686590.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/inference_results/smoke-scaling-2p63M-s42-step10610_GOOG_4686590"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/inference_results/smoke-scaling-2p63M-s42-step10610_GOOG_4686590/data_real",
        "display_path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/inference_results/smoke-scaling-2p63M-s42-step10610_GOOG_4686590/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/lob_pipeline/inference_results/smoke-scaling-2p63M-s42-step10610_GOOG_4686590/data_real",
            "selector": null,
            "source_value": 0.271284,
            "report_record_id": "canonical-scores:37512c45-68fd-814a-accb-f4889b973e71",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8100-a6e1-f305968ba3ce",
      "source_row_id": "37712c45-68fd-8100-a6e1-f305968ba3ce",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8100a6e1f305968ba3ce",
      "run": "24tok-j2633975-s65073-L100sim",
      "stock": null,
      "model_size": null,
      "run_family": "24tok",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.243188,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.243188",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.176284,
      "reported_ks_context": 0.121759,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-05T12:20:22Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2118",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-05T12:20:22Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-05T12:20:22Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-05T12:20:22Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.243188,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8100-a6e1-f305968ba3ce",
        "class": "G20",
        "model_evaluation": "24tok-j2633975-s65073-L100sim",
        "wws21_mean": 0.243188,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-05T12:20:22Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-j2633975-s65073-L100sim_GOOG_3338916/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-j2633975-s65073-L100sim_GOOG_3338916/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-05T12:20:22Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8100a6e1f305968ba3ce",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 681,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-j2633975-s65073-L100sim_GOOG_3338916.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975 (step 65073)",
            "checkpoint_step": 65073,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-j2633975-s65073-L100sim_GOOG_3338916.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-j2633975-s65073-L100sim_GOOG_3338916\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 65073         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262799+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.121665+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-j2633975-s65073-L100sim_GOOG_3338916.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-j2633975-s65073-L100sim_GOOG_3338916\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 65073         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-j2633975-s65073-L100sim_GOOG_3338916.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-j2633975-s65073-L100sim_GOOG_3338916.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-j2633975-s65073-L100sim_GOOG_3338916"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-j2633975-s65073-L100sim_GOOG_3338916/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-j2633975-s65073-L100sim_GOOG_3338916/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-j2633975-s65073-L100sim_GOOG_3338916/data_real.tar",
            "selector": null,
            "source_value": 0.243188,
            "report_record_id": "canonical-scores:37712c45-68fd-8100-a6e1-f305968ba3ce",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8100-a711-ee2c97a10047",
      "source_row_id": "37712c45-68fd-8100-a711-ee2c97a10047",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8100a711ee2c97a10047",
      "run": "pareto-360m-adamw-s46847",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.180782,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.180782",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.137458,
      "reported_ks_context": 0.102211,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:10Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8100a711ee2c97a10047",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:10Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:10Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:10Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.180782,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8100-a711-ee2c97a10047",
        "class": "G20",
        "model_evaluation": "pareto-360m-adamw-s46847",
        "wws21_mean": 0.180782,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:10Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s46847_GOOG_3253403/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s46847_GOOG_3253403/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:10Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8100a711ee2c97a10047",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 682,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s46847_GOOG_3253403.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665 (step 46847)",
            "checkpoint_step": 46847,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s46847_GOOG_3253403.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s46847_GOOG_3253403\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 46847         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262928+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.121688+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s46847_GOOG_3253403.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s46847_GOOG_3253403\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 46847         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s46847_GOOG_3253403.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s46847_GOOG_3253403.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s46847_GOOG_3253403"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s46847_GOOG_3253403/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s46847_GOOG_3253403/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s46847_GOOG_3253403/data_real.tar",
            "selector": null,
            "source_value": 0.180782,
            "report_record_id": "canonical-scores:37712c45-68fd-8100-a711-ee2c97a10047",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8101-b854-c7d03526f593",
      "source_row_id": "37712c45-68fd-8101-b854-c7d03526f593",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8101b854c7d03526f593",
      "run": "s2k-d256-s79740",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.110132,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.110132",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.144524,
      "reported_ks_context": 0.092958,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:11Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8101b854c7d03526f593",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:11Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:11Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:11Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.110132,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8101-b854-c7d03526f593",
        "class": "G20",
        "model_evaluation": "s2k-d256-s79740",
        "wws21_mean": 0.110132,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:11Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s79740_GOOG_3848658/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s79740_GOOG_3848658/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:11Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8101b854c7d03526f593",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 683,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596/79740",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s79740_GOOG_3848658.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596 (step 79740)",
            "checkpoint_step": 79740,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s79740_GOOG_3848658.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s79740_GOOG_3848658\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 79740         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.121699+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s79740_GOOG_3848658.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s79740_GOOG_3848658\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 79740         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s79740_GOOG_3848658.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s79740_GOOG_3848658.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-d256-s79740_GOOG_3848658"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s79740_GOOG_3848658/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s79740_GOOG_3848658/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s79740_GOOG_3848658/data_real.tar",
            "selector": null,
            "source_value": 0.110132,
            "report_record_id": "canonical-scores:37712c45-68fd-8101-b854-c7d03526f593",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8102-b164-c36a06f6ec32",
      "source_row_id": "37712c45-68fd-8102-b164-c36a06f6ec32",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8102b164c36a06f6ec32",
      "run": "genlen-ctx500-c500-g1500",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.283469,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.283469",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.216273,
      "reported_ks_context": 0.15337,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:15Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8102b164c36a06f6ec32",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:15Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:15Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:15Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.283469,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8102-b164-c36a06f6ec32",
        "class": "G20",
        "model_evaluation": "genlen-ctx500-c500-g1500",
        "wws21_mean": 0.283469,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:15Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1500_GOOG_3405776/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1500_GOOG_3405776/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:15Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8102b164c36a06f6ec32",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 684,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1500_GOOG_3405776.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1500_GOOG_3405776.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx500-c500-g1500_GOOG_3405776\"         --n_cond_msgs 500         --n_gen_msgs 1500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262945+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.125757+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 1500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1500_GOOG_3405776.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx500-c500-g1500_GOOG_3405776\"         --n_cond_msgs 500         --n_gen_msgs 1500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1500_GOOG_3405776.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1500_GOOG_3405776.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx500-c500-g1500_GOOG_3405776"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1500_GOOG_3405776/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1500_GOOG_3405776/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1500_GOOG_3405776/data_real.tar",
            "selector": null,
            "source_value": 0.283469,
            "report_record_id": "canonical-scores:37712c45-68fd-8102-b164-c36a06f6ec32",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8104-b5b8-c9c690d447a7",
      "source_row_id": "37712c45-68fd-8104-b5b8-c9c690d447a7",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8104b5b8c9c690d447a7",
      "run": "46m-s2k-d256-c1000g250",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.081118,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.081118",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.130363,
      "reported_ks_context": 0.083374,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:57Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8104b5b8c9c690d447a7",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:57Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:57Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:57Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.081118,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8104-b5b8-c9c690d447a7",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-c1000g250",
        "wws21_mean": 0.081118,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:57Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1000g250_GOOG_3864374/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1000g250_GOOG_3864374/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:57Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8104b5b8c9c690d447a7",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 685,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/83310",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1000g250_GOOG_3864374.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 83310)",
            "checkpoint_step": 83310,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1000g250_GOOG_3864374.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-c1000g250_GOOG_3864374\"         --n_cond_msgs 1000         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 83310         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 1000,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1000g250_GOOG_3864374.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-c1000g250_GOOG_3864374\"         --n_cond_msgs 1000         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 83310         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1000g250_GOOG_3864374.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1000g250_GOOG_3864374.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-c1000g250_GOOG_3864374"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1000g250_GOOG_3864374/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1000g250_GOOG_3864374/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1000g250_GOOG_3864374/data_real.tar",
            "selector": null,
            "source_value": 0.081118,
            "report_record_id": "canonical-scores:37712c45-68fd-8104-b5b8-c9c690d447a7",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8105-b07d-f18b59eb1d68",
      "source_row_id": "37712c45-68fd-8105-b07d-f18b59eb1d68",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8105b07df18b59eb1d68",
      "run": "H2-55M-ctx8k-38k",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.209184,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.209184",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.166756,
      "reported_ks_context": 0.108769,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:53Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8105b07df18b59eb1d68",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:53Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:53Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:53Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.209184,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8105-b07d-f18b59eb1d68",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx8k-38k",
        "wws21_mean": 0.209184,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:53Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-38k/scores/scores_uncond_GOOG_H2-55M-ctx8k-38k_integrated_2572655.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-38k_GOOG_2572655/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:53Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-38k_GOOG_2572655/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8105b07df18b59eb1d68",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 686,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "41850",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/41850",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-38k_GOOG_2572655/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/41850/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/41850",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/41850",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "41850",
                "45450",
                "46650",
                "47850",
                "49050",
                "50250",
                "51450",
                "52650",
                "53849",
                "55048",
                "56248",
                "57434",
                "58091",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/41850",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx8k-38k_GOOG_2572655.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-38k_GOOG_2572655\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 41850         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx8k-38k_GOOG_2572655.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-38k/scores/scores_uncond_GOOG_H2-55M-ctx8k-38k_integrated_2572655.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-38k/scores/scores_uncond_GOOG_H2-55M-ctx8k-38k_integrated_2572655.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.20918415012231245,
        "feature_points": {
          "spread": 0.28279119460855184,
          "ask_cancellation_depth": 0.4247236906510351,
          "bid_cancellation_depth": 0.48250129511039974,
          "limit_ask_order_levels": 0.049640179278770916,
          "limit_bid_order_levels": 0.07543824577624085,
          "ask_cancellation_levels": 0.08229100116603459,
          "bid_cancellation_levels": 0.09158384592865396,
          "vol_per_min": 0.15181867929192322,
          "ofi": 0.15117666071973604,
          "ofi_up": 0.10894439684992385,
          "ofi_stay": 0.15388137394137102,
          "orderbook_imbalance": 0.1024822941289095,
          "ofi_down": 0.137703617645228,
          "log_inter_arrival_time": 0.19101091642439144,
          "log_time_to_cancel": 0.6049322462367669,
          "ask_volume_touch": 0.11290283158534427,
          "bid_volume_touch": 0.07939799735019958,
          "ask_volume": 0.074039996111238,
          "bid_volume": 0.1299769881623714,
          "limit_ask_order_depth": 0.43877229442497445,
          "limit_bid_order_depth": 0.46685740717649615
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-38k/scores/scores_uncond_GOOG_H2-55M-ctx8k-38k_integrated_2572655.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.209184,
            "report_record_id": "canonical-scores:37712c45-68fd-8105-b07d-f18b59eb1d68",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8105-b472-f46e8f5eb66f",
      "source_row_id": "37712c45-68fd-8105-b472-f46e8f5eb66f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8105b472f46e8f5eb66f",
      "run": "nsa-o2b-RoPE-ft",
      "stock": null,
      "model_size": null,
      "run_family": "nsa",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.60508,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.60508",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.421791,
      "reported_ks_context": 0.3528,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:49Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8105b472f46e8f5eb66f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:49Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:49Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:49Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.60508,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8105-b472-f46e8f5eb66f",
        "class": "G20",
        "model_evaluation": "nsa-o2b-RoPE-ft",
        "wws21_mean": 0.60508,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:49Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-RoPE-ft_GOOG_3261367/data_real",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-RoPE-ft_GOOG_3261367/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:49Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8105b472f46e8f5eb66f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 687,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2b_rope/checkpoints/j3260448_bmzwcyr3_3260448/15050",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-RoPE-ft_GOOG_3261367.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_rope/checkpoints/j3260448_bmzwcyr3_3260448 (step 15050)",
            "checkpoint_step": 15050,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_rope/checkpoints/j3260448_bmzwcyr3_3260448",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-RoPE-ft_GOOG_3261367.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_rope         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_rope/checkpoints/j3260448_bmzwcyr3_3260448\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-RoPE-ft_GOOG_3261367\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 15050         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.129809+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2b_rope/checkpoints/j3260448_bmzwcyr3_3260448",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_rope/checkpoints/j3260448_bmzwcyr3_3260448"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-RoPE-ft_GOOG_3261367.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_rope         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_rope/checkpoints/j3260448_bmzwcyr3_3260448\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-RoPE-ft_GOOG_3261367\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 15050         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-RoPE-ft_GOOG_3261367.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-RoPE-ft_GOOG_3261367.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/nsa-o2b-RoPE-ft_GOOG_3261367"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-RoPE-ft_GOOG_3261367/data_real",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-RoPE-ft_GOOG_3261367/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-RoPE-ft_GOOG_3261367/data_real",
            "selector": null,
            "source_value": 0.60508,
            "report_record_id": "canonical-scores:37712c45-68fd-8105-b472-f46e8f5eb66f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8106-9ab4-f89afc4d84a9",
      "source_row_id": "37712c45-68fd-8106-9ab4-f89afc4d84a9",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81069ab4f89afc4d84a9",
      "run": "phaseA-r100-a-s28980-c250g250",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.126195,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.126195",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.162333,
      "reported_ks_context": 0.099822,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:04Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81069ab4f89afc4d84a9",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:04Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:04Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:04Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.126195,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8106-9ab4-f89afc4d84a9",
        "class": "G20",
        "model_evaluation": "phaseA-r100-a-s28980-c250g250",
        "wws21_mean": 0.126195,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:04Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:04Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81069ab4f89afc4d84a9",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 688,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c250g250_GOOG_3701972.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348 (step 28980)",
            "checkpoint_step": 28980,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c250g250_GOOG_3701972.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 28980         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.131648+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c250g250_GOOG_3701972.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 28980         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c250g250_GOOG_3701972.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c250g250_GOOG_3701972.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972/data_real.tar",
            "selector": null,
            "source_value": 0.126195,
            "report_record_id": "canonical-scores:37712c45-68fd-8106-9ab4-f89afc4d84a9",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8107-9476-c81e0ca0fb20",
      "source_row_id": "37712c45-68fd-8107-9476-c81e0ca0fb20",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81079476c81e0ca0fb20",
      "run": "s2k-d256-s30750",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.104631,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.104631",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.158704,
      "reported_ks_context": 0.092386,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:25Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81079476c81e0ca0fb20",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:25Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:25Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:25Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.104631,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8107-9476-c81e0ca0fb20",
        "class": "G20",
        "model_evaluation": "s2k-d256-s30750",
        "wws21_mean": 0.104631,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:25Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s30750_GOOG_3835482/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s30750_GOOG_3835482/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:25Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81079476c81e0ca0fb20",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 689,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596/30750",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s30750_GOOG_3835482.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596 (step 30750)",
            "checkpoint_step": 30750,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s30750_GOOG_3835482.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s30750_GOOG_3835482\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 30750         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.121699+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s30750_GOOG_3835482.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s30750_GOOG_3835482\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 30750         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s30750_GOOG_3835482.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s30750_GOOG_3835482.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-d256-s30750_GOOG_3835482"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s30750_GOOG_3835482/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s30750_GOOG_3835482/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s30750_GOOG_3835482/data_real.tar",
            "selector": null,
            "source_value": 0.104631,
            "report_record_id": "canonical-scores:37712c45-68fd-8107-9476-c81e0ca0fb20",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8109-96e7-fec72b32d7e9",
      "source_row_id": "37712c45-68fd-8109-96e7-fec72b32d7e9",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810996e7fec72b32d7e9",
      "run": "phaseB-m3-14m-s2000-d256-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.086012,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.086012",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.147085,
      "reported_ks_context": 0.092295,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:56Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810996e7fec72b32d7e9",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:56Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:56Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:56Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.086012,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8109-96e7-fec72b32d7e9",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s2000-d256-b-avg-c250g250",
        "wws21_mean": 0.086012,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:56Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:56Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810996e7fec72b32d7e9",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 690,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937/log_rank0.txt",
            "evidence_line": 645,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704065_hpxxr5mb_3704065",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704065_hpxxr5mb_3704065\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704065_hpxxr5mb_3704065 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937/log_rank0.txt",
                "evidence_line": 647,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065",
            "availability_checked_at": "2026-09-12T06:13:15.374097+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704065_hpxxr5mb_3704065\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c250g250_GOOG_3713937/data_real.tar",
            "selector": null,
            "source_value": 0.086012,
            "report_record_id": "canonical-scores:37712c45-68fd-8109-96e7-fec72b32d7e9",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-810a-98de-cd6ee17ab8ab",
      "source_row_id": "37712c45-68fd-810a-98de-cd6ee17ab8ab",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810a98decd6ee17ab8ab",
      "run": "G10-l3k-raw-arm",
      "stock": null,
      "model_size": null,
      "run_family": "G10",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.615816,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.615816",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.40639,
      "reported_ks_context": 0.37965,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:08Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810a98decd6ee17ab8ab",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:08Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:08Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:08Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.615816,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-810a-98de-cd6ee17ab8ab",
        "class": "G20",
        "model_evaluation": "G10-l3k-raw-arm",
        "wws21_mean": 0.615816,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:08Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-raw-arm_GOOG_3512717/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-raw-arm_GOOG_3512717/data_real.tar;/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-raw-arm_GOOG_3508930/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:08Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 2048 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810a98decd6ee17ab8ab",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 691,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169/132203",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-raw-arm_GOOG_3512717.out",
            "evidence_line": 8,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169 (step 132203)",
            "checkpoint_step": 132203,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-raw-arm_GOOG_3512717.out",
                "evidence_line": 22,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l3k-raw-arm_GOOG_3512717\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 2048         --batch_size 64         --checkpoint_step 132203         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.131932+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-raw-arm_GOOG_3512717.out",
            "line": 22,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l3k-raw-arm_GOOG_3512717\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 2048         --batch_size 64         --checkpoint_step 132203         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-raw-arm_GOOG_3512717.out",
            "line": 14,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-raw-arm_GOOG_3512717.out",
            "line": 19,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l3k-raw-arm_GOOG_3512717"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-raw-arm_GOOG_3512717/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-raw-arm_GOOG_3512717/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-raw-arm_GOOG_3512717/data_real.tar",
            "selector": null,
            "source_value": 0.615816,
            "report_record_id": "canonical-scores:37712c45-68fd-810a-98de-cd6ee17ab8ab",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-810c-8be5-dc5ba38a4b1a",
      "source_row_id": "37712c45-68fd-810c-8be5-dc5ba38a4b1a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810c8be5dc5ba38a4b1a",
      "run": "ticks-mamba2-soup5",
      "stock": null,
      "model_size": null,
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.193035,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.193035",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.148444,
      "reported_ks_context": 0.097811,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:30Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810c8be5dc5ba38a4b1a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:30Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:30Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:30Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.193035,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-810c-8be5-dc5ba38a4b1a",
        "class": "G20",
        "model_evaluation": "ticks-mamba2-soup5",
        "wws21_mean": 0.193035,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:30Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-mamba2-soup5/scores/scores_uncond_GOOG_ticks-mamba2-soup5_integrated_3278224.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:30Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 3136 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810c8be5dc5ba38a4b1a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 692,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933/log_rank0.txt",
            "evidence_line": 534,
            "evidence_text": "[Checkpoint] Loading msgpack checkpoint step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5/0",
              "availability": "exists",
              "entries": [
                "metadata",
                "state.msgpack"
              ],
              "root_entries": [
                "0",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5/0",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup5_GOOG_3278224.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup5_GOOG_3278224.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup5_GOOG_3278224.out",
            "line": 132,
            "text": "  Loading: scores_uncond_GOOG_ticks-mamba2-soup5_integrated_3278224.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup5_GOOG_2917933.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup5_GOOG_2917933.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup5_GOOG_3278224.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup5_GOOG_3278224.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup5_GOOG_3278224.out",
            "line": 132,
            "text": "  Loading: scores_uncond_GOOG_ticks-mamba2-soup5_integrated_3278224.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-mamba2-soup5/scores/scores_uncond_GOOG_ticks-mamba2-soup5_integrated_3278224.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-mamba2-soup5/scores/scores_uncond_GOOG_ticks-mamba2-soup5_integrated_3278224.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19303512190888364,
        "feature_points": {
          "vol_per_min": 0.10843922872583028,
          "ofi_up": 0.059467416175028716,
          "ofi_down": 0.050532434342648064,
          "limit_ask_order_ticks": 0.30444723826223563,
          "bid_cancellation_depth": 0.36831380919001366,
          "ask_volume_touch": 0.05001978697560061,
          "spread": 0.1715077652518594,
          "limit_bid_order_depth": 0.3425434864614655,
          "ask_volume": 0.08702573124541642,
          "bid_cancellation_ticks": 0.3677503422960196,
          "limit_bid_order_ticks": 0.3425823152861204,
          "ask_cancellation_depth": 0.32344254756132057,
          "ask_cancellation_ticks": 0.3227472510460292,
          "limit_ask_order_depth": 0.3044480033616066,
          "log_time_to_cancel": 0.3704955829907012,
          "bid_volume_touch": 0.032708121404073545,
          "bid_volume": 0.065672999173649,
          "orderbook_imbalance": 0.09151530266917014,
          "ofi": 0.0894693924207729,
          "log_inter_arrival_time": 0.10950742763322019,
          "ofi_stay": 0.09110137761377542
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-mamba2-soup5/scores/scores_uncond_GOOG_ticks-mamba2-soup5_integrated_3278224.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.193035,
            "report_record_id": "canonical-scores:37712c45-68fd-810c-8be5-dc5ba38a4b1a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-810c-99c9-c88fac2daa1d",
      "source_row_id": "37712c45-68fd-810c-99c9-c88fac2daa1d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810c99c9c88fac2daa1d",
      "run": "phaseB-m3-14m-s1000-d128-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.170289,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.170289",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.195607,
      "reported_ks_context": 0.120316,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:47Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810c99c9c88fac2daa1d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:47Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:47Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:47Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.170289,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-810c-99c9-c88fac2daa1d",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s1000-d128-b-avg-c250g250",
        "wws21_mean": 0.170289,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:47Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:47Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810c99c9c88fac2daa1d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 693,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929/log_rank0.txt",
            "evidence_line": 618,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703794_b5hnvj1k_3703794",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703794_b5hnvj1k_3703794\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703794_b5hnvj1k_3703794 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929/log_rank0.txt",
                "evidence_line": 620,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794",
            "availability_checked_at": "2026-09-12T06:13:15.367418+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703794_b5hnvj1k_3703794\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c250g250_GOOG_3713929/data_real.tar",
            "selector": null,
            "source_value": 0.170289,
            "report_record_id": "canonical-scores:37712c45-68fd-810c-99c9-c88fac2daa1d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-810c-a781-de5ca22ec418",
      "source_row_id": "37712c45-68fd-810c-a781-de5ca22ec418",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810ca781de5ca22ec418",
      "run": "phaseB-m3-14m-s2000-d128-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.188382,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.188382",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.187796,
      "reported_ks_context": 0.120129,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:23Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810ca781de5ca22ec418",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:23Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:23Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:23Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.188382,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-810c-a781-de5ca22ec418",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s2000-d128-b-avg-c250g250",
        "wws21_mean": 0.188382,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:23Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:23Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810ca781de5ca22ec418",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 694,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935/log_rank0.txt",
            "evidence_line": 609,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704413_xgloeib5_3704413",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704413_xgloeib5_3704413\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704413_xgloeib5_3704413 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935/log_rank0.txt",
                "evidence_line": 611,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413",
            "availability_checked_at": "2026-09-12T06:13:15.376162+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704413_xgloeib5_3704413\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c250g250_GOOG_3713935/data_real.tar",
            "selector": null,
            "source_value": 0.188382,
            "report_record_id": "canonical-scores:37712c45-68fd-810c-a781-de5ca22ec418",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-810c-ab80-cd61866499c1",
      "source_row_id": "37712c45-68fd-810c-ab80-cd61866499c1",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810cab80cd61866499c1",
      "run": "s2k-cd-s107330",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.105212,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.105212",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.144764,
      "reported_ks_context": 0.094435,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:07Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810cab80cd61866499c1",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:07Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:07Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:07Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.105212,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-810c-ab80-cd61866499c1",
        "class": "G20",
        "model_evaluation": "s2k-cd-s107330",
        "wws21_mean": 0.105212,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:07Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s107330_GOOG_3864039/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s107330_GOOG_3864039/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:07Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810cab80cd61866499c1",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 695,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/107330",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s107330_GOOG_3864039.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 107330)",
            "checkpoint_step": 107330,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s107330_GOOG_3864039.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s107330_GOOG_3864039\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 107330         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s107330_GOOG_3864039.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s107330_GOOG_3864039\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 107330         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s107330_GOOG_3864039.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s107330_GOOG_3864039.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s107330_GOOG_3864039"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s107330_GOOG_3864039/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s107330_GOOG_3864039/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s107330_GOOG_3864039/data_real.tar",
            "selector": null,
            "source_value": 0.105212,
            "report_record_id": "canonical-scores:37712c45-68fd-810c-ab80-cd61866499c1",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-810d-91ef-f492a2583e8e",
      "source_row_id": "37712c45-68fd-810d-91ef-f492a2583e8e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810d91eff492a2583e8e",
      "run": "pareto-55m-s44634",
      "stock": null,
      "model_size": "55M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.213169,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.213169",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.166008,
      "reported_ks_context": 0.12762,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:01Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810d91eff492a2583e8e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:01Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:01Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:01Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.213169,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-810d-91ef-f492a2583e8e",
        "class": "G20",
        "model_evaluation": "pareto-55m-s44634",
        "wws21_mean": 0.213169,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:01Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s44634_GOOG_3253390/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s44634_GOOG_3253390/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:01Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810d91eff492a2583e8e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 696,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099/44634",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s44634_GOOG_3253390.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099 (step 44634)",
            "checkpoint_step": 44634,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s44634_GOOG_3253390.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s44634_GOOG_3253390\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 44634         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.132988+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s44634_GOOG_3253390.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s44634_GOOG_3253390\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 44634         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s44634_GOOG_3253390.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s44634_GOOG_3253390.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s44634_GOOG_3253390"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s44634_GOOG_3253390/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s44634_GOOG_3253390/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s44634_GOOG_3253390/data_real.tar",
            "selector": null,
            "source_value": 0.213169,
            "report_record_id": "canonical-scores:37712c45-68fd-810d-91ef-f492a2583e8e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-810d-b288-db30d71fddc8",
      "source_row_id": "37712c45-68fd-810d-b288-db30d71fddc8",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810db288db30d71fddc8",
      "run": "46m-s2k-d256-c1750g250",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.093543,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.093543",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.132209,
      "reported_ks_context": 0.084751,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:15Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810db288db30d71fddc8",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:15Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:15Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:15Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.093543,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-810d-b288-db30d71fddc8",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-c1750g250",
        "wws21_mean": 0.093543,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:15Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1750g250_GOOG_3875944/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1750g250_GOOG_3875944/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:15Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810db288db30d71fddc8",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 697,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/83310",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1750g250_GOOG_3875944.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 83310)",
            "checkpoint_step": 83310,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1750g250_GOOG_3875944.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-c1750g250_GOOG_3875944\"         --n_cond_msgs 1750         --n_gen_msgs 250         --n_sequences 1024         --batch_size 16         --checkpoint_step 83310         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 1750,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1750g250_GOOG_3875944.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-c1750g250_GOOG_3875944\"         --n_cond_msgs 1750         --n_gen_msgs 250         --n_sequences 1024         --batch_size 16         --checkpoint_step 83310         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1750g250_GOOG_3875944.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-c1750g250_GOOG_3875944.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-c1750g250_GOOG_3875944"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1750g250_GOOG_3875944/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1750g250_GOOG_3875944/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-c1750g250_GOOG_3875944/data_real.tar",
            "selector": null,
            "source_value": 0.093543,
            "report_record_id": "canonical-scores:37712c45-68fd-810d-b288-db30d71fddc8",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-810e-a21b-e1a003370236",
      "source_row_id": "37712c45-68fd-810e-a21b-e1a003370236",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810ea21be1a003370236",
      "run": "pareto-120m-s135458",
      "stock": null,
      "model_size": "120M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.174186,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.174186",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.127256,
      "reported_ks_context": 0.098065,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:15Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810ea21be1a003370236",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:15Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:15Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:15Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.174186,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-810e-a21b-e1a003370236",
        "class": "G20",
        "model_evaluation": "pareto-120m-s135458",
        "wws21_mean": 0.174186,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:15Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s135458_GOOG_3253399/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s135458_GOOG_3253399/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:15Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810ea21be1a003370236",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 698,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s135458_GOOG_3253399.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440 (step 135458)",
            "checkpoint_step": 135458,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s135458_GOOG_3253399.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s135458_GOOG_3253399\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 135458         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.134047+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s135458_GOOG_3253399.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s135458_GOOG_3253399\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 135458         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s135458_GOOG_3253399.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s135458_GOOG_3253399.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s135458_GOOG_3253399"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s135458_GOOG_3253399/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s135458_GOOG_3253399/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s135458_GOOG_3253399/data_real.tar",
            "selector": null,
            "source_value": 0.174186,
            "report_record_id": "canonical-scores:37712c45-68fd-810e-a21b-e1a003370236",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-810f-9bb4-ce4fb14d1816",
      "source_row_id": "37712c45-68fd-810f-9bb4-ce4fb14d1816",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd810f9bb4ce4fb14d1816",
      "run": "genlen-ctx4k-c500-g1500",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.236204,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.236204",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.190343,
      "reported_ks_context": 0.139212,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:08Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd810f9bb4ce4fb14d1816",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:08Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:08Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:08Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.236204,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-810f-9bb4-ce4fb14d1816",
        "class": "G20",
        "model_evaluation": "genlen-ctx4k-c500-g1500",
        "wws21_mean": 0.236204,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:08Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1500_GOOG_3405786/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1500_GOOG_3405786/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:08Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd810f9bb4ce4fb14d1816",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 699,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1500_GOOG_3405786.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1500_GOOG_3405786.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g1500_GOOG_3405786\"         --n_cond_msgs 500         --n_gen_msgs 1500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262982+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134305+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 1500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1500_GOOG_3405786.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g1500_GOOG_3405786\"         --n_cond_msgs 500         --n_gen_msgs 1500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1500_GOOG_3405786.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1500_GOOG_3405786.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g1500_GOOG_3405786"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1500_GOOG_3405786/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1500_GOOG_3405786/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1500_GOOG_3405786/data_real.tar",
            "selector": null,
            "source_value": 0.236204,
            "report_record_id": "canonical-scores:37712c45-68fd-810f-9bb4-ce4fb14d1816",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8110-8a98-ee85bc265426",
      "source_row_id": "37712c45-68fd-8110-8a98-ee85bc265426",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81108a98ee85bc265426",
      "run": "H1_75M_step10828",
      "stock": null,
      "model_size": "75M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.23357,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.23357",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.177264,
      "reported_ks_context": 0.116467,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:35Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81108a98ee85bc265426",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:35Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:35Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:35Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.23357,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8110-8a98-ee85bc265426",
        "class": "G20",
        "model_evaluation": "H1_75M_step10828",
        "wws21_mean": 0.23357,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:35Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_75M_step10828/scores/scores_uncond_GOOG_H1_75M_step10828_integrated_2566034.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1_75M_step10828_GOOG_2566034/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:35Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1_75M_step10828_GOOG_2566034/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81108a98ee85bc265426",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 700,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "10828",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916/10828",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1_75M_step10828_GOOG_2566034/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916/10828/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916/10828",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916/10828",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "10828",
                "12718",
                "2037",
                "4242",
                "6446",
                "8640",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916/10828",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1_75M_step10828_GOOG_2566034.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1_75M_step10828_GOOG_2566034\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 10828         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1_75M_step10828_GOOG_2566034.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_75M_step10828/scores/scores_uncond_GOOG_H1_75M_step10828_integrated_2566034.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_75M_step10828/scores/scores_uncond_GOOG_H1_75M_step10828_integrated_2566034.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.23357011262803176,
        "feature_points": {
          "spread": 0.4418594488233584,
          "ask_cancellation_depth": 0.4081902716633023,
          "bid_cancellation_depth": 0.4126333679512141,
          "limit_ask_order_levels": 0.09370624761668296,
          "limit_bid_order_levels": 0.09103445961523464,
          "ask_cancellation_levels": 0.166046228686363,
          "bid_cancellation_levels": 0.1713386160877158,
          "vol_per_min": 0.16177816755684749,
          "ofi": 0.09796554407068238,
          "ofi_up": 0.1588751798482646,
          "ofi_stay": 0.09729124189077475,
          "orderbook_imbalance": 0.10014886864939852,
          "ofi_down": 0.1844666556616634,
          "log_inter_arrival_time": 0.07620395189848497,
          "log_time_to_cancel": 0.5930151551745146,
          "ask_volume_touch": 0.2177335045432096,
          "bid_volume_touch": 0.2038828401008209,
          "ask_volume": 0.17455341030026766,
          "bid_volume": 0.2346687111060366,
          "limit_ask_order_depth": 0.41167473145013256,
          "limit_bid_order_depth": 0.4079057624936976
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_75M_step10828/scores/scores_uncond_GOOG_H1_75M_step10828_integrated_2566034.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.23357,
            "report_record_id": "canonical-scores:37712c45-68fd-8110-8a98-ee85bc265426",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8110-9397-cf5bdfe98f07",
      "source_row_id": "37712c45-68fd-8110-9397-cf5bdfe98f07",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81109397cf5bdfe98f07",
      "run": "mamba2-77M-soup10",
      "stock": null,
      "model_size": "77M",
      "run_family": "mamba2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.137953,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.137953",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.105795,
      "reported_ks_context": 0.075076,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:57Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81109397cf5bdfe98f07",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:57Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:57Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:57Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.137953,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8110-9397-cf5bdfe98f07",
        "class": "G20",
        "model_evaluation": "mamba2-77M-soup10",
        "wws21_mean": 0.137953,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:57Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:57Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81109397cf5bdfe98f07",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 701,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10/0",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.134324+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
            "selector": null,
            "source_value": 0.137953,
            "report_record_id": "canonical-scores:37712c45-68fd-8110-9397-cf5bdfe98f07",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8110-ad35-d19df5e5fa08",
      "source_row_id": "37712c45-68fd-8110-ad35-d19df5e5fa08",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8110ad35d19df5e5fa08",
      "run": "soup-gdn-rand",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.153155,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.153155",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.124847,
      "reported_ks_context": 0.076987,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:01Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8110ad35d19df5e5fa08",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:01Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:01Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:01Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.153155,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8110-ad35-d19df5e5fa08",
        "class": "G20",
        "model_evaluation": "soup-gdn-rand",
        "wws21_mean": 0.153155,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:01Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:01Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8110ad35d19df5e5fa08",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 702,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-rand_GOOG_2888137.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-rand_GOOG_2888137.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262991+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134585+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-rand_GOOG_2888137.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-rand_GOOG_2888137.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-rand_GOOG_2888137.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137/data_real.tar",
            "selector": null,
            "source_value": 0.153155,
            "report_record_id": "canonical-scores:37712c45-68fd-8110-ad35-d19df5e5fa08",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8110-b744-c862116f598e",
      "source_row_id": "37712c45-68fd-8110-b744-c862116f598e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8110b744c862116f598e",
      "run": "phaseA-r100-a-s28980-c500g500",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.109916,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.109916",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.170769,
      "reported_ks_context": 0.103073,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:52Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8110b744c862116f598e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:52Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:52Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:52Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.109916,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8110-b744-c862116f598e",
        "class": "G20",
        "model_evaluation": "phaseA-r100-a-s28980-c500g500",
        "wws21_mean": 0.109916,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:52Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:52Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8110b744c862116f598e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 703,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c500g500_GOOG_3701973.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348 (step 28980)",
            "checkpoint_step": 28980,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c500g500_GOOG_3701973.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 28980         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.131648+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c500g500_GOOG_3701973.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 28980         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c500g500_GOOG_3701973.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c500g500_GOOG_3701973.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973/data_real.tar",
            "selector": null,
            "source_value": 0.109916,
            "report_record_id": "canonical-scores:37712c45-68fd-8110-b744-c862116f598e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8110-bf3a-e646b1889951",
      "source_row_id": "37712c45-68fd-8110-bf3a-e646b1889951",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8110bf3ae646b1889951",
      "run": "muon-s49680",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.18021,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.18021",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.133691,
      "reported_ks_context": 0.096881,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:52Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8110bf3ae646b1889951",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:52Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:52Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:52Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.18021,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8110-bf3a-e646b1889951",
        "class": "G20",
        "model_evaluation": "muon-s49680",
        "wws21_mean": 0.18021,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:52Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s49680_GOOG_2868871/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s49680_GOOG_2868871/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:52Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8110bf3ae646b1889951",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 704,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s49680_GOOG_2868871.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131 (step 49680)",
            "checkpoint_step": 49680,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s49680_GOOG_2868871.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s49680_GOOG_2868871\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 49680         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262995+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134605+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s49680_GOOG_2868871.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s49680_GOOG_2868871\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 49680         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s49680_GOOG_2868871.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s49680_GOOG_2868871.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-s49680_GOOG_2868871"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s49680_GOOG_2868871/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s49680_GOOG_2868871/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s49680_GOOG_2868871/data_real.tar",
            "selector": null,
            "source_value": 0.18021,
            "report_record_id": "canonical-scores:37712c45-68fd-8110-bf3a-e646b1889951",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8112-8856-ce099216db00",
      "source_row_id": "37712c45-68fd-8112-8856-ce099216db00",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81128856ce099216db00",
      "run": "H2-55M-ctx8k-11k",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.197806,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.197806",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.160731,
      "reported_ks_context": 0.097752,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:19Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81128856ce099216db00",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:19Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:19Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:19Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.197806,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8112-8856-ce099216db00",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx8k-11k",
        "wws21_mean": 0.197806,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:19Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-11k/scores/scores_uncond_GOOG_H2-55M-ctx8k-11k_integrated_2572653.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-11k_GOOG_2572653/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:19Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-11k_GOOG_2572653/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81128856ce099216db00",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 705,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "11880",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2504810_x9xvs3wz_2504810/11880",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2504810_x9xvs3wz_2504810",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-11k_GOOG_2572653/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2504810_x9xvs3wz_2504810/11880/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2504810_x9xvs3wz_2504810/11880",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2504810_x9xvs3wz_2504810/11880",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "1105",
                "11880",
                "15470",
                "16668",
                "17865",
                "19062",
                "20260",
                "21458",
                "22656",
                "23854",
                "25051",
                "26248",
                "27445",
                "4700",
                "7095",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2504810_x9xvs3wz_2504810/11880",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx8k-11k_GOOG_2572653.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2504810_x9xvs3wz_2504810\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-11k_GOOG_2572653\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 11880         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx8k-11k_GOOG_2572653.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-11k/scores/scores_uncond_GOOG_H2-55M-ctx8k-11k_integrated_2572653.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-11k/scores/scores_uncond_GOOG_H2-55M-ctx8k-11k_integrated_2572653.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19780553922129424,
        "feature_points": {
          "spread": 0.47748475648154304,
          "ask_cancellation_depth": 0.3952195362981289,
          "bid_cancellation_depth": 0.43605619032777065,
          "limit_ask_order_levels": 0.07036715340909942,
          "limit_bid_order_levels": 0.0629280974355688,
          "ask_cancellation_levels": 0.10217992725447472,
          "bid_cancellation_levels": 0.094501155330518,
          "vol_per_min": 0.1048119090302759,
          "ofi": 0.06664359782440317,
          "ofi_up": 0.08424697473817451,
          "ofi_stay": 0.06770488132389506,
          "orderbook_imbalance": 0.0957526419625965,
          "ofi_down": 0.0913506617042677,
          "log_inter_arrival_time": 0.06294368318456331,
          "log_time_to_cancel": 0.5154843653065428,
          "ask_volume_touch": 0.13426074302695268,
          "bid_volume_touch": 0.1455899624769708,
          "ask_volume": 0.10896192888178953,
          "bid_volume": 0.19274500429670186,
          "limit_ask_order_depth": 0.4078414207598816,
          "limit_bid_order_depth": 0.43684173259305975
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-11k/scores/scores_uncond_GOOG_H2-55M-ctx8k-11k_integrated_2572653.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.197806,
            "report_record_id": "canonical-scores:37712c45-68fd-8112-8856-ce099216db00",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8112-8d99-e9aa22f0ff38",
      "source_row_id": "37712c45-68fd-8112-8d99-e9aa22f0ff38",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81128d99e9aa22f0ff38",
      "run": "14m-s2k-d512",
      "stock": null,
      "model_size": "14M",
      "run_family": "14m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.105309,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.105309",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.146866,
      "reported_ks_context": 0.096763,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:47Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81128d99e9aa22f0ff38",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:47Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:47Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:47Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.105309,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8112-8d99-e9aa22f0ff38",
        "class": "G20",
        "model_evaluation": "14m-s2k-d512",
        "wws21_mean": 0.105309,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:47Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d512_GOOG_3873792/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d512_GOOG_3873792/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:47Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81128d99e9aa22f0ff38",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 706,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865377_ac8ww5nm_3865377/41150",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d512_GOOG_3873792.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865377_ac8ww5nm_3865377 (step 41150)",
            "checkpoint_step": 41150,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865377_ac8ww5nm_3865377",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d512_GOOG_3873792.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865377_ac8ww5nm_3865377\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/14m-s2k-d512_GOOG_3873792\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 41150         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.134615+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865377_ac8ww5nm_3865377",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865377_ac8ww5nm_3865377"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d512_GOOG_3873792.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865377_ac8ww5nm_3865377\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/14m-s2k-d512_GOOG_3873792\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 41150         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d512_GOOG_3873792.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d512_GOOG_3873792.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/14m-s2k-d512_GOOG_3873792"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d512_GOOG_3873792/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d512_GOOG_3873792/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d512_GOOG_3873792/data_real.tar",
            "selector": null,
            "source_value": 0.105309,
            "report_record_id": "canonical-scores:37712c45-68fd-8112-8d99-e9aa22f0ff38",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8112-9653-eb88c132d1ee",
      "source_row_id": "37712c45-68fd-8112-9653-eb88c132d1ee",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81129653eb88c132d1ee",
      "run": "adamw-32n-s31614",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.155385,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.155385",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.116706,
      "reported_ks_context": 0.084133,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:02Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81129653eb88c132d1ee",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:02Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:02Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:02Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.155385,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8112-9653-eb88c132d1ee",
        "class": "G20",
        "model_evaluation": "adamw-32n-s31614",
        "wws21_mean": 0.155385,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:02Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s31614_GOOG_2871392/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s31614_GOOG_2871392/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:02Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81129653eb88c132d1ee",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 707,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s31614_GOOG_2871392.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 31614)",
            "checkpoint_step": 31614,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s31614_GOOG_2871392.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s31614_GOOG_2871392\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 31614         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s31614_GOOG_2871392.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s31614_GOOG_2871392\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 31614         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s31614_GOOG_2871392.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s31614_GOOG_2871392.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s31614_GOOG_2871392"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s31614_GOOG_2871392/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s31614_GOOG_2871392/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s31614_GOOG_2871392/data_real.tar",
            "selector": null,
            "source_value": 0.155385,
            "report_record_id": "canonical-scores:37712c45-68fd-8112-9653-eb88c132d1ee",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8113-ad1d-ef5c130c1114",
      "source_row_id": "37712c45-68fd-8113-ad1d-ef5c130c1114",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8113ad1def5c130c1114",
      "run": "pareto-55m-s93224",
      "stock": null,
      "model_size": "55M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.193926,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.193926",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.143635,
      "reported_ks_context": 0.111327,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:29Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8113ad1def5c130c1114",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:29Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:29Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:29Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.193926,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8113-ad1d-ef5c130c1114",
        "class": "G20",
        "model_evaluation": "pareto-55m-s93224",
        "wws21_mean": 0.193926,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:29Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s93224_GOOG_3253392/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s93224_GOOG_3253392/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:29Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8113ad1def5c130c1114",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 708,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s93224_GOOG_3253392.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279 (step 93224)",
            "checkpoint_step": 93224,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s93224_GOOG_3253392.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s93224_GOOG_3253392\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 93224         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.136054+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s93224_GOOG_3253392.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s93224_GOOG_3253392\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 93224         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s93224_GOOG_3253392.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s93224_GOOG_3253392.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s93224_GOOG_3253392"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s93224_GOOG_3253392/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s93224_GOOG_3253392/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s93224_GOOG_3253392/data_real.tar",
            "selector": null,
            "source_value": 0.193926,
            "report_record_id": "canonical-scores:37712c45-68fd-8113-ad1d-ef5c130c1114",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8113-b929-c9a96d8d7005",
      "source_row_id": "37712c45-68fd-8113-b929-c9a96d8d7005",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8113b929c9a96d8d7005",
      "run": "ctx4k-cond2000",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.158752,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.158752",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.127325,
      "reported_ks_context": 0.089725,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:33Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8113b929c9a96d8d7005",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:33Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:33Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:33Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.158752,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8113-b929-c9a96d8d7005",
        "class": "G20",
        "model_evaluation": "ctx4k-cond2000",
        "wws21_mean": 0.158752,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:33Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond2000/scores/scores_uncond_GOOG_ctx4k-cond2000_integrated_2828762.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond2000_GOOG_2828762/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:33Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond2000_GOOG_2828762/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8113b929c9a96d8d7005",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 709,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond2000_GOOG_2828762/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 2000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond2000_GOOG_2828762.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond2000_GOOG_2828762\"         --n_cond_msgs 2000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond2000_GOOG_2828762.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond2000/scores/scores_uncond_GOOG_ctx4k-cond2000_integrated_2828762.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond2000/scores/scores_uncond_GOOG_ctx4k-cond2000_integrated_2828762.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15875220600819887,
        "feature_points": {
          "ofi_up": 0.05707667082777247,
          "bid_volume_touch": 0.09193700922682557,
          "bid_cancellation_levels": 0.029539750323376884,
          "ask_cancellation_levels": 0.06368317874075148,
          "limit_bid_order_levels": 0.015082814639208053,
          "limit_ask_order_levels": 0.06794927679454207,
          "ofi_down": 0.08155030803227767,
          "spread": 0.30528049282262754,
          "ask_volume_touch": 0.11153634104941693,
          "vol_per_min": 0.06776246879346538,
          "limit_ask_order_depth": 0.3042168776201116,
          "ask_cancellation_depth": 0.35659551163634484,
          "limit_bid_order_depth": 0.3284106255689051,
          "ofi_stay": 0.06564111101909671,
          "bid_cancellation_depth": 0.33526254358499086,
          "ask_volume": 0.24838308062951678,
          "ofi": 0.06372470927876982,
          "log_time_to_cancel": 0.36737057265960316,
          "orderbook_imbalance": 0.09142692349026363,
          "bid_volume": 0.23456831765614575,
          "log_inter_arrival_time": 0.04679774177816394
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond2000/scores/scores_uncond_GOOG_ctx4k-cond2000_integrated_2828762.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.158752,
            "report_record_id": "canonical-scores:37712c45-68fd-8113-b929-c9a96d8d7005",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8114-ad73-d66ce04696cb",
      "source_row_id": "37712c45-68fd-8114-ad73-d66ce04696cb",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8114ad73d66ce04696cb",
      "run": "muon-ga4-s19288",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 1.645965,
      "source_value_field": "Wasserstein",
      "source_value_text": "1.645965",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.529327,
      "reported_ks_context": 0.335522,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:21Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8114ad73d66ce04696cb",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:21Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:21Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:21Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 1.645965,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8114-ad73-d66ce04696cb",
        "class": "G20",
        "model_evaluation": "muon-ga4-s19288",
        "wws21_mean": 1.645965,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:21Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s19288_GOOG_2869389/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s19288_GOOG_2869389/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:21Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8114ad73d66ce04696cb",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 710,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s19288_GOOG_2869389.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716 (step 19288)",
            "checkpoint_step": 19288,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s19288_GOOG_2869389.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-ga4-s19288_GOOG_2869389\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 19288         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263011+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136759+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s19288_GOOG_2869389.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-ga4-s19288_GOOG_2869389\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 19288         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s19288_GOOG_2869389.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s19288_GOOG_2869389.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-ga4-s19288_GOOG_2869389"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s19288_GOOG_2869389/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s19288_GOOG_2869389/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s19288_GOOG_2869389/data_real.tar",
            "selector": null,
            "source_value": 1.645965,
            "report_record_id": "canonical-scores:37712c45-68fd-8114-ad73-d66ce04696cb",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8115-ba48-ce51f2e6249e",
      "source_row_id": "37712c45-68fd-8115-ba48-ce51f2e6249e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8115ba48ce51f2e6249e",
      "run": "R1-s4k-d256-raw-s30190",
      "stock": null,
      "model_size": null,
      "run_family": "R1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.113865,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.113865",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.152055,
      "reported_ks_context": 0.096522,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:25Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8115ba48ce51f2e6249e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:25Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:25Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:25Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.113865,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8115-ba48-ce51f2e6249e",
        "class": "G20",
        "model_evaluation": "R1-s4k-d256-raw-s30190",
        "wws21_mean": 0.113865,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:25Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s30190_GOOG_3745459/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s30190_GOOG_3745459/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:25Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8115ba48ce51f2e6249e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 711,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096/30190",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s30190_GOOG_3745459.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096 (step 30190)",
            "checkpoint_step": 30190,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s30190_GOOG_3745459.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s30190_GOOG_3745459\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 30190         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.136864+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s30190_GOOG_3745459.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s30190_GOOG_3745459\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 30190         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s30190_GOOG_3745459.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s30190_GOOG_3745459.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s30190_GOOG_3745459"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s30190_GOOG_3745459/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s30190_GOOG_3745459/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s30190_GOOG_3745459/data_real.tar",
            "selector": null,
            "source_value": 0.113865,
            "report_record_id": "canonical-scores:37712c45-68fd-8115-ba48-ce51f2e6249e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8116-8ce8-f7d63881f341",
      "source_row_id": "37712c45-68fd-8116-8ce8-f7d63881f341",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81168ce8f7d63881f341",
      "run": "ctx4k-cond1500",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.165965,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.165965",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.124813,
      "reported_ks_context": 0.091651,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:41Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81168ce8f7d63881f341",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:41Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:41Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:41Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.165965,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8116-8ce8-f7d63881f341",
        "class": "G20",
        "model_evaluation": "ctx4k-cond1500",
        "wws21_mean": 0.165965,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:41Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond1500/scores/scores_uncond_GOOG_ctx4k-cond1500_integrated_2828761.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond1500_GOOG_2828761/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:41Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond1500_GOOG_2828761/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81168ce8f7d63881f341",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 712,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond1500_GOOG_2828761/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond1500_GOOG_2828761.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond1500_GOOG_2828761\"         --n_cond_msgs 1500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond1500_GOOG_2828761.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond1500/scores/scores_uncond_GOOG_ctx4k-cond1500_integrated_2828761.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond1500/scores/scores_uncond_GOOG_ctx4k-cond1500_integrated_2828761.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1659647903267684,
        "feature_points": {
          "ask_cancellation_depth": 0.32839893234531387,
          "ask_volume": 0.22992711979034128,
          "ofi_up": 0.06036115042370195,
          "ofi_down": 0.0940131317086296,
          "bid_cancellation_levels": 0.03416408918194687,
          "limit_bid_order_levels": 0.016135821448514893,
          "vol_per_min": 0.14426891277144482,
          "ask_cancellation_levels": 0.05272291683766916,
          "limit_ask_order_levels": 0.07587793844454375,
          "spread": 0.2919281659055364,
          "bid_volume": 0.33881596312321866,
          "ofi": 0.0437570022500432,
          "limit_ask_order_depth": 0.29085200796194405,
          "bid_cancellation_depth": 0.35159526483597653,
          "log_time_to_cancel": 0.35688144733421,
          "limit_bid_order_depth": 0.34411980401861336,
          "ask_volume_touch": 0.15366277091690314,
          "orderbook_imbalance": 0.1046669161374192,
          "ofi_stay": 0.043911101344157565,
          "log_inter_arrival_time": 0.046267501050986114,
          "bid_volume_touch": 0.08293263903102194
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond1500/scores/scores_uncond_GOOG_ctx4k-cond1500_integrated_2828761.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.165965,
            "report_record_id": "canonical-scores:37712c45-68fd-8116-8ce8-f7d63881f341",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8117-9770-c9e5f3374081",
      "source_row_id": "37712c45-68fd-8117-9770-c9e5f3374081",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81179770c9e5f3374081",
      "run": "soup-ctxt2k-rand",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.162824,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.162824",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.127426,
      "reported_ks_context": 0.091479,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:20Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81179770c9e5f3374081",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:20Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:20Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:20Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.162824,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8117-9770-c9e5f3374081",
        "class": "G20",
        "model_evaluation": "soup-ctxt2k-rand",
        "wws21_mean": 0.162824,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:20Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:20Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81179770c9e5f3374081",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 713,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt2k-rand_GOOG_2888141.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt2k-rand_GOOG_2888141.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263020+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.139444+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt2k-rand_GOOG_2888141.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt2k-rand_GOOG_2888141.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt2k-rand_GOOG_2888141.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141/data_real.tar",
            "selector": null,
            "source_value": 0.162824,
            "report_record_id": "canonical-scores:37712c45-68fd-8117-9770-c9e5f3374081",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8117-a8aa-ef71ba85f1ec",
      "source_row_id": "37712c45-68fd-8117-a8aa-ef71ba85f1ec",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8117a8aaef71ba85f1ec",
      "run": "phaseB-m3-14m-s1000-d128-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.186163,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.186163",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.210686,
      "reported_ks_context": 0.130339,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:12Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8117a8aaef71ba85f1ec",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:12Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:12Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:12Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.186163,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8117-a8aa-ef71ba85f1ec",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s1000-d128-b-avg-c500g500",
        "wws21_mean": 0.186163,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:12Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:12Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8117a8aaef71ba85f1ec",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 714,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930/log_rank0.txt",
            "evidence_line": 626,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703794_b5hnvj1k_3703794",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703794_b5hnvj1k_3703794\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703794_b5hnvj1k_3703794 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930/log_rank0.txt",
                "evidence_line": 628,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794",
            "availability_checked_at": "2026-09-12T06:13:15.367418+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703794_b5hnvj1k_3703794/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703794_b5hnvj1k_3703794\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d128-b-avg-c500g500_GOOG_3713930/data_real.tar",
            "selector": null,
            "source_value": 0.186163,
            "report_record_id": "canonical-scores:37712c45-68fd-8117-a8aa-ef71ba85f1ec",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8118-a45c-ee9c62117adb",
      "source_row_id": "37712c45-68fd-8118-a45c-ee9c62117adb",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8118a45cee9c62117adb",
      "run": "muon-ga4-s21711",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.159291,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.159291",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.121153,
      "reported_ks_context": 0.085607,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:18Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8118a45cee9c62117adb",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:18Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:18Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:18Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.159291,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8118-a45c-ee9c62117adb",
        "class": "G20",
        "model_evaluation": "muon-ga4-s21711",
        "wws21_mean": 0.159291,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:18Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s21711_GOOG_2869390/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s21711_GOOG_2869390/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:18Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8118a45cee9c62117adb",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 715,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s21711_GOOG_2869390.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716 (step 21711)",
            "checkpoint_step": 21711,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s21711_GOOG_2869390.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-ga4-s21711_GOOG_2869390\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 21711         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263011+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136759+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s21711_GOOG_2869390.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2823716_p4qoxjln_2823716\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-ga4-s21711_GOOG_2869390\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 21711         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s21711_GOOG_2869390.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-ga4-s21711_GOOG_2869390.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-ga4-s21711_GOOG_2869390"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s21711_GOOG_2869390/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s21711_GOOG_2869390/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-ga4-s21711_GOOG_2869390/data_real.tar",
            "selector": null,
            "source_value": 0.159291,
            "report_record_id": "canonical-scores:37712c45-68fd-8118-a45c-ee9c62117adb",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8118-a88b-c9f05e01dd7c",
      "source_row_id": "37712c45-68fd-8118-a88b-c9f05e01dd7c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8118a88bc9f05e01dd7c",
      "run": "s2k-d256-c2000",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.109737,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.109737",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.146459,
      "reported_ks_context": 0.099662,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8118a88bc9f05e01dd7c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.109737,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8118-a88b-c9f05e01dd7c",
        "class": "G20",
        "model_evaluation": "s2k-d256-c2000",
        "wws21_mean": 0.109737,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-c2000_GOOG_3762573/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-c2000_GOOG_3762573/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8118a88bc9f05e01dd7c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 716,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-c2000_GOOG_3762573.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065 (step 999999)",
            "checkpoint_step": 999999,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-c2000_GOOG_3762573.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-c2000_GOOG_3762573\"         --n_cond_msgs 2000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:13:15.374097+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 2000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-c2000_GOOG_3762573.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-c2000_GOOG_3762573\"         --n_cond_msgs 2000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-c2000_GOOG_3762573.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-c2000_GOOG_3762573.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-d256-c2000_GOOG_3762573"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-c2000_GOOG_3762573/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-c2000_GOOG_3762573/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-c2000_GOOG_3762573/data_real.tar",
            "selector": null,
            "source_value": 0.109737,
            "report_record_id": "canonical-scores:37712c45-68fd-8118-a88b-c9f05e01dd7c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8118-acb8-db149348ef6d",
      "source_row_id": "37712c45-68fd-8118-acb8-db149348ef6d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8118acb8db149348ef6d",
      "run": "ctx4k-cond3000",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.156445,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.156445",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.124762,
      "reported_ks_context": 0.089214,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:53Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8118acb8db149348ef6d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:53Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:53Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:53Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.156445,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8118-acb8-db149348ef6d",
        "class": "G20",
        "model_evaluation": "ctx4k-cond3000",
        "wws21_mean": 0.156445,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:53Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond3000/scores/scores_uncond_GOOG_ctx4k-cond3000_integrated_2828764.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond3000_GOOG_2828764/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:53Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond3000_GOOG_2828764/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8118acb8db149348ef6d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 717,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond3000_GOOG_2828764/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 3000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond3000_GOOG_2828764.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond3000_GOOG_2828764\"         --n_cond_msgs 3000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond3000_GOOG_2828764.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond3000/scores/scores_uncond_GOOG_ctx4k-cond3000_integrated_2828764.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond3000/scores/scores_uncond_GOOG_ctx4k-cond3000_integrated_2828764.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15644473241363507,
        "feature_points": {
          "spread": 0.2810688751289308,
          "limit_bid_order_depth": 0.34551164619634794,
          "limit_ask_order_depth": 0.2784127724037269,
          "limit_ask_order_levels": 0.057800300802584956,
          "ask_cancellation_levels": 0.03869823294926893,
          "ask_cancellation_depth": 0.3095290458450942,
          "vol_per_min": 0.15701704195391822,
          "ofi_up": 0.04589283494307858,
          "limit_bid_order_levels": 0.018820524386963047,
          "bid_cancellation_levels": 0.028451835338818326,
          "ofi_down": 0.05575522599283705,
          "bid_cancellation_depth": 0.3318561465144395,
          "ask_volume": 0.18905275098956106,
          "bid_volume": 0.2959201546136055,
          "log_time_to_cancel": 0.37045813532842714,
          "ask_volume_touch": 0.13261649439230816,
          "bid_volume_touch": 0.08263803176314655,
          "ofi_stay": 0.05739572439571111,
          "orderbook_imbalance": 0.10418509975954415,
          "ofi": 0.05566606425250112,
          "log_inter_arrival_time": 0.048592442735523246
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond3000/scores/scores_uncond_GOOG_ctx4k-cond3000_integrated_2828764.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.156445,
            "report_record_id": "canonical-scores:37712c45-68fd-8118-acb8-db149348ef6d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8119-8195-e9daa2f62f27",
      "source_row_id": "37712c45-68fd-8119-8195-e9daa2f62f27",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81198195e9daa2f62f27",
      "run": "phaseB-m3-14m-s500-d128-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.161707,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.161707",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.163367,
      "reported_ks_context": 0.110591,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:41Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81198195e9daa2f62f27",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:41Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:41Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:41Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.161707,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8119-8195-e9daa2f62f27",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s500-d128-b-avg-c500g500",
        "wws21_mean": 0.161707,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:41Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:41Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81198195e9daa2f62f27",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 718,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948/log_rank0.txt",
            "evidence_line": 612,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703791_os296u12_3703791",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703791_os296u12_3703791\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703791_os296u12_3703791 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948/log_rank0.txt",
                "evidence_line": 614,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791",
            "availability_checked_at": "2026-09-12T06:13:15.366404+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703791_os296u12_3703791\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c500g500_GOOG_3713948/data_real.tar",
            "selector": null,
            "source_value": 0.161707,
            "report_record_id": "canonical-scores:37712c45-68fd-8119-8195-e9daa2f62f27",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8119-a083-c87e20eb5840",
      "source_row_id": "37712c45-68fd-8119-a083-c87e20eb5840",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8119a083c87e20eb5840",
      "run": "H6_MoE_551M_step11380",
      "stock": null,
      "model_size": "551M",
      "run_family": "H6",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.186473,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.186473",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.151028,
      "reported_ks_context": 0.094283,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:33Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8119a083c87e20eb5840",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:33Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:33Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:33Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.186473,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8119-a083-c87e20eb5840",
        "class": "G20",
        "model_evaluation": "H6_MoE_551M_step11380",
        "wws21_mean": 0.186473,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:33Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6_MoE_551M_step11380/scores/scores_uncond_GOOG_H6_MoE_551M_step11380_shard16_integrated_2566037.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H6_MoE_551M_step11380_GOOG_2566037/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:33Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H6_MoE_551M_step11380_GOOG_2566037/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8119a083c87e20eb5840",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 719,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "11380",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2504809_xemc7z3r_2504809/11380",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2504809_xemc7z3r_2504809",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H6_MoE_551M_step11380_GOOG_2566037/log_rank0.txt",
            "evidence_line": 855,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2504809_xemc7z3r_2504809/11380/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2504809_xemc7z3r_2504809/11380",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2504809_xemc7z3r_2504809/11380",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "10556",
                "11380",
                "12200",
                "13018",
                "4005",
                "5639",
                "6461",
                "7280",
                "8097",
                "8922",
                "9741",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2504809_xemc7z3r_2504809/11380",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H6_MoE_551M_step11380_GOOG_2566037.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2504809_xemc7z3r_2504809\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H6_MoE_551M_step11380_GOOG_2566037\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 11380         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H6_MoE_551M_step11380_GOOG_2566037.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6_MoE_551M_step11380/scores/scores_uncond_GOOG_H6_MoE_551M_step11380_shard16_integrated_2566037.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6_MoE_551M_step11380/scores/scores_uncond_GOOG_H6_MoE_551M_step11380_integrated_2566037.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.18647343184559087,
        "feature_points": {
          "spread": 0.18374978093406336,
          "ask_cancellation_depth": 0.3531835644499106,
          "bid_cancellation_depth": 0.3778613490819028,
          "limit_ask_order_levels": 0.043843501857970285,
          "limit_bid_order_levels": 0.03247170274250741,
          "ask_cancellation_levels": 0.11591564060434394,
          "bid_cancellation_levels": 0.08643707879840239,
          "vol_per_min": 0.17151463485452256,
          "ofi": 0.12945036513746572,
          "ofi_up": 0.07367636281127832,
          "ofi_stay": 0.13167808223091618,
          "orderbook_imbalance": 0.09931458917491014,
          "ofi_down": 0.051876085110171874,
          "log_inter_arrival_time": 0.11510212014917004,
          "log_time_to_cancel": 0.6586649802271357,
          "ask_volume_touch": 0.14475479357085227,
          "bid_volume_touch": 0.11979104309282834,
          "ask_volume": 0.12582481535870943,
          "bid_volume": 0.13114552956711126,
          "limit_ask_order_depth": 0.372772320224325,
          "limit_bid_order_depth": 0.396913728778911
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6_MoE_551M_step11380/scores/scores_uncond_GOOG_H6_MoE_551M_step11380_integrated_2566037.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.186473,
            "report_record_id": "canonical-scores:37712c45-68fd-8119-a083-c87e20eb5840",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-811a-a8d9-ed66aeb9189d",
      "source_row_id": "37712c45-68fd-811a-a8d9-ed66aeb9189d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd811aa8d9ed66aeb9189d",
      "run": "phaseA-r075-a-s30960-c500g500",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.150111,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.150111",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.1619,
      "reported_ks_context": 0.106935,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:50Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd811aa8d9ed66aeb9189d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:50Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:50Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:50Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.150111,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-811a-a8d9-ed66aeb9189d",
        "class": "G20",
        "model_evaluation": "phaseA-r075-a-s30960-c500g500",
        "wws21_mean": 0.150111,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:50Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c500g500_GOOG_3701971/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c500g500_GOOG_3701971/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:50Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd811aa8d9ed66aeb9189d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 720,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347/30960",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c500g500_GOOG_3701971.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347 (step 30960)",
            "checkpoint_step": 30960,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c500g500_GOOG_3701971.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r075-a-s30960-c500g500_GOOG_3701971\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 30960         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.141730+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c500g500_GOOG_3701971.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r075-a-s30960-c500g500_GOOG_3701971\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 30960         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c500g500_GOOG_3701971.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c500g500_GOOG_3701971.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-r075-a-s30960-c500g500_GOOG_3701971"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c500g500_GOOG_3701971/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c500g500_GOOG_3701971/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c500g500_GOOG_3701971/data_real.tar",
            "selector": null,
            "source_value": 0.150111,
            "report_record_id": "canonical-scores:37712c45-68fd-811a-a8d9-ed66aeb9189d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-811b-a1ea-ebabde1a7a6e",
      "source_row_id": "37712c45-68fd-811b-a1ea-ebabde1a7a6e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd811ba1eaebabde1a7a6e",
      "run": "J1-anchor100-61k",
      "stock": null,
      "model_size": null,
      "run_family": "J1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.679298,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.679298",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.573296,
      "reported_ks_context": 0.480687,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:53Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd811ba1eaebabde1a7a6e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:53Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:53Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:53Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.679298,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-811b-a1ea-ebabde1a7a6e",
        "class": "G20",
        "model_evaluation": "J1-anchor100-61k",
        "wws21_mean": 0.679298,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:53Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor100-61k/scores/scores_uncond_GOOG_J1-anchor100-61k_integrated_2654628.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor100-61k_GOOG_2654628/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:53Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor100-61k_GOOG_2654628/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd811ba1eaebabde1a7a6e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 721,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "60980",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976/60980",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor100-61k_GOOG_2654628/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976/60980/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976/60980",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976/60980",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976/60980",
            "availability": "not_found",
            "path_scope": "checkpoint_step",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976/60980",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_J1-anchor100-61k_GOOG_2654628.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/J1-anchor100-61k_GOOG_2654628\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 60980         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_J1-anchor100-61k_GOOG_2654628.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "incomplete_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor100-61k/scores/scores_uncond_GOOG_J1-anchor100-61k_integrated_2654628.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor100-61k/scores/scores_uncond_GOOG_J1-anchor100-61k_integrated_2654628.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 20,
        "nonfinite_features": [
          "orderbook_imbalance"
        ],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "reviewed_nonfinite_feature_counts",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor100-61k/scores/scores_uncond_GOOG_J1-anchor100-61k_integrated_2654628.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.679298,
            "report_record_id": "canonical-scores:37712c45-68fd-811b-a1ea-ebabde1a7a6e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "The reviewed 21-feature score file contains nonfinite feature values. The reported source score is retained; it is not a verified complete finite 21-feature mean. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-811b-b63a-caec7e845db8",
      "source_row_id": "37712c45-68fd-811b-b63a-caec7e845db8",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd811bb63acaec7e845db8",
      "run": "N1-base-s63328-L100sim",
      "stock": null,
      "model_size": null,
      "run_family": "N1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.221763,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.221763",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.190682,
      "reported_ks_context": 0.142671,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:13Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd811bb63acaec7e845db8",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:13Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:13Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:13Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.221763,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-811b-b63a-caec7e845db8",
        "class": "G20",
        "model_evaluation": "N1-base-s63328-L100sim",
        "wws21_mean": 0.221763,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:13Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim_GOOG_3234655/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim_GOOG_3234655/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:13Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd811bb63acaec7e845db8",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 722,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391/63328",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim_GOOG_3234655.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391 (step 63328)",
            "checkpoint_step": 63328,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim_GOOG_3234655.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1-base-s63328-L100sim_GOOG_3234655\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 63328         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.143072+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim_GOOG_3234655.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1-base-s63328-L100sim_GOOG_3234655\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 63328         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim_GOOG_3234655.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim_GOOG_3234655.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1-base-s63328-L100sim_GOOG_3234655"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim_GOOG_3234655/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim_GOOG_3234655/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim_GOOG_3234655/data_real.tar",
            "selector": null,
            "source_value": 0.221763,
            "report_record_id": "canonical-scores:37712c45-68fd-811b-b63a-caec7e845db8",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-811c-ada2-e4344bc54717",
      "source_row_id": "37712c45-68fd-811c-ada2-e4344bc54717",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd811cada2e4344bc54717",
      "run": "phaseB-m3-14m-s1000-d256-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.111487,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.111487",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.162143,
      "reported_ks_context": 0.09877,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:08Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd811cada2e4344bc54717",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:08Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:08Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:08Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.111487,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-811c-ada2-e4344bc54717",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s1000-d256-b-avg-c250g250",
        "wws21_mean": 0.111487,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:08Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:08Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd811cada2e4344bc54717",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 723,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931/log_rank0.txt",
            "evidence_line": 651,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704412_bmp7l150_3704412",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704412_bmp7l150_3704412\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704412_bmp7l150_3704412 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931/log_rank0.txt",
                "evidence_line": 653,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412",
            "availability_checked_at": "2026-09-12T06:13:15.375258+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704412_bmp7l150_3704412\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c250g250_GOOG_3713931/data_real.tar",
            "selector": null,
            "source_value": 0.111487,
            "report_record_id": "canonical-scores:37712c45-68fd-811c-ada2-e4344bc54717",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-811e-9b94-fc0860bd0258",
      "source_row_id": "37712c45-68fd-811e-9b94-fc0860bd0258",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd811e9b94fc0860bd0258",
      "run": "ticks-ctx1k-rand",
      "stock": null,
      "model_size": null,
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.222637,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.222637",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.17747,
      "reported_ks_context": 0.122195,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:14Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd811e9b94fc0860bd0258",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:14Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:14Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:14Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.222637,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-811e-9b94-fc0860bd0258",
        "class": "G20",
        "model_evaluation": "ticks-ctx1k-rand",
        "wws21_mean": 0.222637,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:14Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx1k-rand/scores/scores_uncond_GOOG_ticks-ctx1k-rand_integrated_3278228.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:14Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd811e9b94fc0860bd0258",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 724,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139/log_rank0.txt",
            "evidence_line": 634,
            "evidence_text": "[Checkpoint] Loading step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx1k-rand_GOOG_3278228.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx1k-rand_GOOG_3278228.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx1k-rand_GOOG_3278228.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-ctx1k-rand_integrated_3278228.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt1k-rand_GOOG_2888139.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt1k-rand_GOOG_2888139.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx1k-rand_GOOG_3278228.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx1k-rand_GOOG_3278228.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx1k-rand_GOOG_3278228.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-ctx1k-rand_integrated_3278228.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx1k-rand/scores/scores_uncond_GOOG_ticks-ctx1k-rand_integrated_3278228.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx1k-rand/scores/scores_uncond_GOOG_ticks-ctx1k-rand_integrated_3278228.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.2226371929715173,
        "feature_points": {
          "ask_cancellation_depth": 0.30290450365183386,
          "limit_bid_order_ticks": 0.3310571913386863,
          "ofi_down": 0.11932135216243596,
          "bid_cancellation_ticks": 0.31802099653828086,
          "limit_bid_order_depth": 0.3313487902325168,
          "limit_ask_order_ticks": 0.3128228994968075,
          "ask_cancellation_ticks": 0.30209021369709566,
          "ask_volume_touch": 0.11009131452708185,
          "limit_ask_order_depth": 0.3129776467405501,
          "ofi_up": 0.1428619344966874,
          "spread": 0.2806594731123289,
          "vol_per_min": 0.13058805952904134,
          "ofi": 0.09739297892977836,
          "bid_cancellation_depth": 0.3187372574387142,
          "log_time_to_cancel": 0.40161369104922184,
          "orderbook_imbalance": 0.10313109307811663,
          "ofi_stay": 0.09661147398378978,
          "bid_volume_touch": 0.11844151406805319,
          "bid_volume": 0.2890149750109269,
          "ask_volume": 0.23336598625709307,
          "log_inter_arrival_time": 0.022327707062822544
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx1k-rand/scores/scores_uncond_GOOG_ticks-ctx1k-rand_integrated_3278228.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.222637,
            "report_record_id": "canonical-scores:37712c45-68fd-811e-9b94-fc0860bd0258",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-811e-b198-f3144704d1eb",
      "source_row_id": "37712c45-68fd-811e-b198-f3144704d1eb",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd811eb198f3144704d1eb",
      "run": "genlen-ctx1k-c500-g1500",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.30278,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.30278",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.238886,
      "reported_ks_context": 0.165671,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd811eb198f3144704d1eb",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.30278,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-811e-b198-f3144704d1eb",
        "class": "G20",
        "model_evaluation": "genlen-ctx1k-c500-g1500",
        "wws21_mean": 0.30278,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1500_GOOG_3405779/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1500_GOOG_3405779/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd811eb198f3144704d1eb",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 725,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1500_GOOG_3405779.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1500_GOOG_3405779.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx1k-c500-g1500_GOOG_3405779\"         --n_cond_msgs 500         --n_gen_msgs 1500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263039+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.145435+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 1500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1500_GOOG_3405779.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx1k-c500-g1500_GOOG_3405779\"         --n_cond_msgs 500         --n_gen_msgs 1500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1500_GOOG_3405779.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1500_GOOG_3405779.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx1k-c500-g1500_GOOG_3405779"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1500_GOOG_3405779/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1500_GOOG_3405779/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1500_GOOG_3405779/data_real.tar",
            "selector": null,
            "source_value": 0.30278,
            "report_record_id": "canonical-scores:37712c45-68fd-811e-b198-f3144704d1eb",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-811f-8c2f-fe1fe9086285",
      "source_row_id": "37712c45-68fd-811f-8c2f-fe1fe9086285",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd811f8c2ffe1fe9086285",
      "run": "adamw_j2715665_52900",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.186114,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.186114",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.143689,
      "reported_ks_context": 0.091474,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-10T20:56:04Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2066",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-10T20:56:04Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-10T20:56:04Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-10T20:56:04Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.186114,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-811f-8c2f-fe1fe9086285",
        "class": "G20",
        "model_evaluation": "adamw_j2715665_52900",
        "wws21_mean": 0.186114,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-10T20:56:04Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_adamw_j2715665_52900/scores/scores_uncond_GOOG_adamw_j2715665_52900_integrated_2773111.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/adamw_j2715665_52900_GOOG_2773111/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-10T20:56:04Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/adamw_j2715665_52900_GOOG_2773111/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd811f8c2ffe1fe9086285",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 726,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "52900",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/adamw_j2715665_52900_GOOG_2773111/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=52900 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665/52900",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665/52900",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw_j2715665_52900_GOOG_2773111.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/adamw_j2715665_52900_GOOG_2773111\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 52900         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw_j2715665_52900_GOOG_2773111.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_adamw_j2715665_52900/scores/scores_uncond_GOOG_adamw_j2715665_52900_integrated_2773111.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_adamw_j2715665_52900/scores/scores_uncond_GOOG_adamw_j2715665_52900_integrated_2773111.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1861141017848056,
        "feature_points": {
          "bid_cancellation_depth": 0.4104787954521032,
          "ask_cancellation_levels": 0.08141927226829361,
          "ask_cancellation_depth": 0.3702517367177173,
          "ofi_down": 0.08283491686616563,
          "bid_cancellation_levels": 0.1217596410644556,
          "spread": 0.27096130634106963,
          "limit_ask_order_levels": 0.05970585148638098,
          "bid_volume_touch": 0.06495388021746758,
          "limit_bid_order_levels": 0.09634182787419925,
          "ofi_up": 0.07317840124994472,
          "limit_ask_order_depth": 0.38429940804173374,
          "vol_per_min": 0.08017719024222797,
          "limit_bid_order_depth": 0.4025880587814004,
          "log_time_to_cancel": 0.4474265614224553,
          "orderbook_imbalance": 0.10106362659121365,
          "ask_volume_touch": 0.07648520029987857,
          "ask_volume": 0.15278383123547118,
          "ofi": 0.1510423074222696,
          "bid_volume": 0.19404515878399087,
          "ofi_stay": 0.1525717316925396,
          "log_inter_arrival_time": 0.13402743342993959
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_adamw_j2715665_52900/scores/scores_uncond_GOOG_adamw_j2715665_52900_integrated_2773111.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.186114,
            "report_record_id": "canonical-scores:37712c45-68fd-811f-8c2f-fe1fe9086285",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-811f-bb57-f93b9d394d83",
      "source_row_id": "37712c45-68fd-811f-bb57-f93b9d394d83",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd811fbb57f93b9d394d83",
      "run": "phaseB-m3-14m-h128-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.216679,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.216679",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.210808,
      "reported_ks_context": 0.12323,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:15Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd811fbb57f93b9d394d83",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:15Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:15Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:15Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.216679,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-811f-bb57-f93b9d394d83",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-h128-b-avg-c500g500",
        "wws21_mean": 0.216679,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:15Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:15Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd811fbb57f93b9d394d83",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 727,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928/log_rank0.txt",
            "evidence_line": 603,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703803_39ifyzj9_3703803",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703803_39ifyzj9_3703803\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703803_39ifyzj9_3703803 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928/log_rank0.txt",
                "evidence_line": 605,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803",
            "availability_checked_at": "2026-09-12T06:13:15.371914+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703803_39ifyzj9_3703803\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c500g500_GOOG_3713928/data_real.tar",
            "selector": null,
            "source_value": 0.216679,
            "report_record_id": "canonical-scores:37712c45-68fd-811f-bb57-f93b9d394d83",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8120-860f-f0a4d2fb7f1f",
      "source_row_id": "37712c45-68fd-8120-860f-f0a4d2fb7f1f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8120860ff0a4d2fb7f1f",
      "run": "pareto-120m-s12586",
      "stock": null,
      "model_size": "120M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.239791,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.239791",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.17384,
      "reported_ks_context": 0.132213,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:29Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8120860ff0a4d2fb7f1f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:29Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:29Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:29Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.239791,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8120-860f-f0a4d2fb7f1f",
        "class": "G20",
        "model_evaluation": "pareto-120m-s12586",
        "wws21_mean": 0.239791,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:29Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s12586_GOOG_3253394/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s12586_GOOG_3253394/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:29Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8120860ff0a4d2fb7f1f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 728,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260/12586",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s12586_GOOG_3253394.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260 (step 12586)",
            "checkpoint_step": 12586,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s12586_GOOG_3253394.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s12586_GOOG_3253394\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 12586         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.145456+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s12586_GOOG_3253394.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s12586_GOOG_3253394\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 12586         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s12586_GOOG_3253394.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s12586_GOOG_3253394.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s12586_GOOG_3253394"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s12586_GOOG_3253394/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s12586_GOOG_3253394/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s12586_GOOG_3253394/data_real.tar",
            "selector": null,
            "source_value": 0.239791,
            "report_record_id": "canonical-scores:37712c45-68fd-8120-860f-f0a4d2fb7f1f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8120-89f2-d13151e3b310",
      "source_row_id": "37712c45-68fd-8120-89f2-d13151e3b310",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812089f2d13151e3b310",
      "run": "genlen-ctx4k-c500-g1000",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.22304,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.22304",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.177399,
      "reported_ks_context": 0.124459,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:33Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812089f2d13151e3b310",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:33Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:33Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:33Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.22304,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8120-89f2-d13151e3b310",
        "class": "G20",
        "model_evaluation": "genlen-ctx4k-c500-g1000",
        "wws21_mean": 0.22304,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:33Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1000_GOOG_3405785/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1000_GOOG_3405785/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:33Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812089f2d13151e3b310",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 729,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1000_GOOG_3405785.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1000_GOOG_3405785.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g1000_GOOG_3405785\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262982+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134305+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 1000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1000_GOOG_3405785.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g1000_GOOG_3405785\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1000_GOOG_3405785.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g1000_GOOG_3405785.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g1000_GOOG_3405785"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1000_GOOG_3405785/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1000_GOOG_3405785/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g1000_GOOG_3405785/data_real.tar",
            "selector": null,
            "source_value": 0.22304,
            "report_record_id": "canonical-scores:37712c45-68fd-8120-89f2-d13151e3b310",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8123-8491-f5e4f21043c3",
      "source_row_id": "37712c45-68fd-8123-8491-f5e4f21043c3",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81238491f5e4f21043c3",
      "run": "adamw_j2731367_33522",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.136207,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.136207",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.111003,
      "reported_ks_context": 0.075144,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-11T10:56:04Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2045",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-11T10:56:04Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-11T10:56:04Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-11T10:56:04Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.136207,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8123-8491-f5e4f21043c3",
        "class": "G20",
        "model_evaluation": "adamw_j2731367_33522",
        "wws21_mean": 0.136207,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-11T10:56:04Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_adamw_j2731367_33522/scores/scores_uncond_GOOG_adamw_j2731367_33522_integrated_2779129.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/adamw_j2731367_33522_GOOG_2779129/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-11T10:56:04Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/adamw_j2731367_33522_GOOG_2779129/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81238491f5e4f21043c3",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 730,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "33522",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/adamw_j2731367_33522_GOOG_2779129/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=33522 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367/33522",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367/33522",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw_j2731367_33522_GOOG_2779129.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/adamw_j2731367_33522_GOOG_2779129\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 33522         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw_j2731367_33522_GOOG_2779129.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_adamw_j2731367_33522/scores/scores_uncond_GOOG_adamw_j2731367_33522_integrated_2779129.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_adamw_j2731367_33522/scores/scores_uncond_GOOG_adamw_j2731367_33522_integrated_2779129.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1362068307713916,
        "feature_points": {
          "limit_bid_order_levels": 0.08418459777420101,
          "ofi_down": 0.07471944549977176,
          "bid_cancellation_levels": 0.04957910614864766,
          "limit_ask_order_depth": 0.3195385991476943,
          "ofi_up": 0.04356004678413411,
          "ask_cancellation_depth": 0.29935025113023167,
          "ofi_stay": 0.09043984533023948,
          "vol_per_min": 0.08163864428856646,
          "bid_cancellation_depth": 0.3179567888673796,
          "limit_ask_order_levels": 0.090428310432477,
          "log_inter_arrival_time": 0.032267535761210185,
          "ask_cancellation_levels": 0.0757162218625598,
          "bid_volume_touch": 0.03696800795059835,
          "limit_bid_order_depth": 0.3366125769095293,
          "spread": 0.15140504156351892,
          "ask_volume": 0.0878033521928129,
          "ask_volume_touch": 0.04775266935642077,
          "log_time_to_cancel": 0.31194272253860567,
          "bid_volume": 0.1330228745428859,
          "ofi": 0.08927304641135969,
          "orderbook_imbalance": 0.10618376170637912
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_adamw_j2731367_33522/scores/scores_uncond_GOOG_adamw_j2731367_33522_integrated_2779129.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.136207,
            "report_record_id": "canonical-scores:37712c45-68fd-8123-8491-f5e4f21043c3",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8123-86eb-fb8dea56d5e2",
      "source_row_id": "37712c45-68fd-8123-86eb-fb8dea56d5e2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812386ebfb8dea56d5e2",
      "run": "phaseB-m3-14m-s4000-d64-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.187418,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.187418",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.163495,
      "reported_ks_context": 0.102335,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:16Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812386ebfb8dea56d5e2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:16Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:16Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:16Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.187418,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8123-86eb-fb8dea56d5e2",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s4000-d64-b-avg-c500g500",
        "wws21_mean": 0.187418,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:16Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:16Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812386ebfb8dea56d5e2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 731,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946/log_rank0.txt",
            "evidence_line": 595,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703799_69cc21ax_3703799",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703799_69cc21ax_3703799\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703799_69cc21ax_3703799 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946/log_rank0.txt",
                "evidence_line": 597,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799",
            "availability_checked_at": "2026-09-12T06:13:15.370080+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703799_69cc21ax_3703799\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c500g500_GOOG_3713946/data_real.tar",
            "selector": null,
            "source_value": 0.187418,
            "report_record_id": "canonical-scores:37712c45-68fd-8123-86eb-fb8dea56d5e2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8124-8e33-c1516a345ff7",
      "source_row_id": "37712c45-68fd-8124-8e33-c1516a345ff7",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81248e33c1516a345ff7",
      "run": "genlen-ctx2k-c500-g2000",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.243289,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.243289",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.227917,
      "reported_ks_context": 0.162144,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:22Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81248e33c1516a345ff7",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:22Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:22Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:22Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.243289,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8124-8e33-c1516a345ff7",
        "class": "G20",
        "model_evaluation": "genlen-ctx2k-c500-g2000",
        "wws21_mean": 0.243289,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:22Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g2000_GOOG_3405783/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g2000_GOOG_3405783/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:22Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81248e33c1516a345ff7",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 732,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g2000_GOOG_3405783.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g2000_GOOG_3405783.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g2000_GOOG_3405783\"         --n_cond_msgs 500         --n_gen_msgs 2000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263020+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.139444+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 2000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g2000_GOOG_3405783.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g2000_GOOG_3405783\"         --n_cond_msgs 500         --n_gen_msgs 2000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g2000_GOOG_3405783.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g2000_GOOG_3405783.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g2000_GOOG_3405783"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g2000_GOOG_3405783/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g2000_GOOG_3405783/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g2000_GOOG_3405783/data_real.tar",
            "selector": null,
            "source_value": 0.243289,
            "report_record_id": "canonical-scores:37712c45-68fd-8124-8e33-c1516a345ff7",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8126-93f5-c3a25e1ce258",
      "source_row_id": "37712c45-68fd-8126-93f5-c3a25e1ce258",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812693f5c3a25e1ce258",
      "run": "j2505279-step93224-L100",
      "stock": null,
      "model_size": null,
      "run_family": "j2505279",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.169048,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.169048",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.128854,
      "reported_ks_context": 0.092177,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-28T02:21:52Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2166",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-28T02:21:52Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-28T02:21:52Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-28T02:21:52Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.169048,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8126-93f5-c3a25e1ce258",
        "class": "G20",
        "model_evaluation": "j2505279-step93224-L100",
        "wws21_mean": 0.169048,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-28T02:21:52Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2505279-step93224-L100/scores/scores_uncond_GOOG_j2505279-step93224-L100_integrated_2756974.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2505279-step93224-L100_GOOG_2756974/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-28T02:21:52Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2505279-step93224-L100_GOOG_2756974/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812693f5c3a25e1ce258",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 733,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "93224",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2505279-step93224-L100_GOOG_2756974/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=93224 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "68935",
                "75535",
                "77739",
                "79951",
                "82165",
                "84376",
                "86581",
                "88783",
                "90981",
                "93182",
                "93224",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505279-step93224-L100_GOOG_2756974.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2505279-step93224-L100_GOOG_2756974\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 93224         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505279-step93224-L100_GOOG_2756974.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2505279-step93224-L100/scores/scores_uncond_GOOG_j2505279-step93224-L100_integrated_2756974.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2505279-step93224-L100/scores/scores_uncond_GOOG_j2505279-step93224-L100_integrated_2756974.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.16904808994057421,
        "feature_points": {
          "limit_ask_order_levels": 0.0530270712792503,
          "limit_bid_order_levels": 0.06452617323853702,
          "ofi_stay": 0.04089904661529285,
          "orderbook_imbalance": 0.09609425789259544,
          "spread": 0.27755591057441914,
          "ofi_up": 0.07232284959225366,
          "vol_per_min": 0.18280556993522437,
          "ask_volume_touch": 0.1186629753651276,
          "bid_cancellation_levels": 0.0859012867307276,
          "ask_cancellation_levels": 0.06975589137606017,
          "ofi_down": 0.07766994243543665,
          "log_time_to_cancel": 0.3763669622785919,
          "bid_cancellation_depth": 0.335692267199377,
          "bid_volume_touch": 0.12601561124177263,
          "ask_cancellation_depth": 0.3307742783697815,
          "limit_ask_order_depth": 0.33649897391335754,
          "limit_bid_order_depth": 0.34607662532518696,
          "bid_volume": 0.2541360624957756,
          "ask_volume": 0.22619666750249512,
          "ofi": 0.040345199402160956,
          "log_inter_arrival_time": 0.03868626598863425
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2505279-step93224-L100/scores/scores_uncond_GOOG_j2505279-step93224-L100_integrated_2756974.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.169048,
            "report_record_id": "canonical-scores:37712c45-68fd-8126-93f5-c3a25e1ce258",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8126-bb57-f3c46558e7d2",
      "source_row_id": "37712c45-68fd-8126-bb57-f3c46558e7d2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8126bb57f3c46558e7d2",
      "run": "j2496072_ti7ek5z5",
      "stock": null,
      "model_size": null,
      "run_family": "j2496072",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.954785,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.954785",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.712974,
      "reported_ks_context": 0.549349,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T10:31:40Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2256",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T10:31:40Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T10:31:40Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T10:31:40Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.954785,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8126-bb57-f3c46558e7d2",
        "class": "G20",
        "model_evaluation": "j2496072_ti7ek5z5",
        "wws21_mean": 0.954785,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T10:31:40Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_shard2_integrated_2643169.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2832163/data_real.tar;/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2643169/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T10:31:40Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2832163/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8126bb57f3c46558e7d2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 734,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "28495",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2643169/log_rank0.txt",
            "evidence_line": 465,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "11337",
                "13247",
                "15160",
                "17061",
                "18965",
                "20871",
                "22773",
                "24680",
                "26591",
                "28495",
                "9435",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Multiple source input candidates were resolved by the exact score filename explicitly listed in the matching integrated-job log. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496072_ti7ek5z5_GOOG_2643169.out",
            "line": 518,
            "text": "  - /lus/lfs1aip2/projects/s5e/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_shard2_integrated_2643169.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 256,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496072_ti7ek5z5_GOOG_2643169.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2643169\"         --n_cond_msgs 256         --n_gen_msgs 500         --n_sequences 1024         --batch_size 16         --checkpoint_step 28495         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496072_ti7ek5z5_GOOG_2643169.out",
            "line": 12,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496072_ti7ek5z5_GOOG_2643169.out",
            "line": 518,
            "text": "  - /lus/lfs1aip2/projects/s5e/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_shard2_integrated_2643169.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "incomplete_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_shard2_integrated_2643169.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_integrated_2643169.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 20,
        "nonfinite_features": [
          "orderbook_imbalance"
        ],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "reviewed_nonfinite_feature_counts",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_integrated_2643169.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.954785,
            "report_record_id": "canonical-scores:37712c45-68fd-8126-bb57-f3c46558e7d2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "The reviewed 21-feature score file contains nonfinite feature values. The reported source score is retained; it is not a verified complete finite 21-feature mean. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8126-bda7-d77ceee09ef2",
      "source_row_id": "37712c45-68fd-8126-bda7-d77ceee09ef2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8126bda7d77ceee09ef2",
      "run": "H1-120M-cond1000-gen1000",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.15263,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.15263",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.120989,
      "reported_ks_context": 0.089866,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:32Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8126bda7d77ceee09ef2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:32Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:32Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:32Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.15263,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8126-bda7-d77ceee09ef2",
        "class": "G20",
        "model_evaluation": "H1-120M-cond1000-gen1000",
        "wws21_mean": 0.15263,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:32Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1000-gen1000/scores/scores_uncond_GOOG_H1-120M-cond1000-gen1000_shard9_integrated_2609516.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond1000-gen1000_GOOG_2609516/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:32Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond1000-gen1000_GOOG_2609516/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8126bda7d77ceee09ef2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 735,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond1000-gen1000_GOOG_2609516/log_rank0.txt",
            "evidence_line": 649,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1000,
        "generation_length": 1000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond1000-gen1000_GOOG_2609516.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-120M-cond1000-gen1000_GOOG_2609516\"         --n_cond_msgs 1000         --n_gen_msgs 1000         --n_sequences 1024         --batch_size 8         --checkpoint_step 135458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond1000-gen1000_GOOG_2609516.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1000-gen1000/scores/scores_uncond_GOOG_H1-120M-cond1000-gen1000_shard9_integrated_2609516.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1000-gen1000/scores/scores_uncond_GOOG_H1-120M-cond1000-gen1000_integrated_2609516.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15263042802484886,
        "feature_points": {
          "spread": 0.3433226849589759,
          "ask_cancellation_depth": 0.3044269497620538,
          "bid_cancellation_depth": 0.307956081497511,
          "limit_ask_order_levels": 0.10649247763484446,
          "limit_bid_order_levels": 0.07841614868501277,
          "ask_cancellation_levels": 0.11901205954861678,
          "bid_cancellation_levels": 0.09877734301619054,
          "vol_per_min": 0.1375864998321433,
          "ofi": 0.05356870292883305,
          "ofi_up": 0.05393616273390445,
          "ofi_stay": 0.05350440155769008,
          "orderbook_imbalance": 0.1164099948664342,
          "ofi_down": 0.05809158783995424,
          "log_inter_arrival_time": 0.038517153617838425,
          "log_time_to_cancel": 0.3289170200606655,
          "ask_volume_touch": 0.10224837841581493,
          "bid_volume_touch": 0.041108660320416726,
          "ask_volume": 0.1618459357690544,
          "bid_volume": 0.10306669518826894,
          "limit_ask_order_depth": 0.28281451315959566,
          "limit_bid_order_depth": 0.315219537128007
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1000-gen1000/scores/scores_uncond_GOOG_H1-120M-cond1000-gen1000_integrated_2609516.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.15263,
            "report_record_id": "canonical-scores:37712c45-68fd-8126-bda7-d77ceee09ef2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8128-9b3f-cc8236d8e914",
      "source_row_id": "37712c45-68fd-8128-9b3f-cc8236d8e914",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81289b3fcc8236d8e914",
      "run": "H2-55M-ctx4k-cond3500",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.124663,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.124663",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.112399,
      "reported_ks_context": 0.073478,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:40Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81289b3fcc8236d8e914",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:40Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:40Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:40Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.124663,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8128-9b3f-cc8236d8e914",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond3500",
        "wws21_mean": 0.124663,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:40Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond3500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond3500_shard13_integrated_2576339.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond3500_GOOG_2576339/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:40Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond3500_GOOG_2576339/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81289b3fcc8236d8e914",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 736,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond3500_GOOG_2576339/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 3500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond3500_GOOG_2576339.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond3500_GOOG_2576339\"         --n_cond_msgs 3500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond3500_GOOG_2576339.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond3500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond3500_shard13_integrated_2576339.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond3500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond3500_integrated_2576339.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.12466280605658567,
        "feature_points": {
          "spread": 0.18268297748333295,
          "ask_cancellation_depth": 0.2926618004587321,
          "bid_cancellation_depth": 0.32564554848121996,
          "limit_ask_order_levels": 0.019021069971254054,
          "limit_bid_order_levels": 0.026585165313671978,
          "ask_cancellation_levels": 0.03641220608693733,
          "bid_cancellation_levels": 0.02451802196459242,
          "vol_per_min": 0.031261176140956075,
          "ofi": 0.02389314567262076,
          "ofi_up": 0.04405649865563537,
          "ofi_stay": 0.023853619253706135,
          "orderbook_imbalance": 0.0908331669395527,
          "ofi_down": 0.08367112948588648,
          "log_inter_arrival_time": 0.04068304576825379,
          "log_time_to_cancel": 0.36565051835443063,
          "ask_volume_touch": 0.08652601320226219,
          "bid_volume_touch": 0.09189268016289617,
          "ask_volume": 0.09737755503150962,
          "bid_volume": 0.13293191221859252,
          "limit_ask_order_depth": 0.26006711955597417,
          "limit_bid_order_depth": 0.3376945569862817
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond3500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond3500_integrated_2576339.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.124663,
            "report_record_id": "canonical-scores:37712c45-68fd-8128-9b3f-cc8236d8e914",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8129-82d3-f21ea3bbb17d",
      "source_row_id": "37712c45-68fd-8129-82d3-f21ea3bbb17d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812982d3f21ea3bbb17d",
      "run": "phaseA-r000-ext-c250g250",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.174764,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.174764",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.163857,
      "reported_ks_context": 0.105315,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:36Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812982d3f21ea3bbb17d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:36Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:36Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:36Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.174764,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8129-82d3-f21ea3bbb17d",
        "class": "G20",
        "model_evaluation": "phaseA-r000-ext-c250g250",
        "wws21_mean": 0.174764,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:36Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r000-ext-c250g250/scores/scores_uncond_GOOG_phaseA-r000-ext-c250g250_integrated_3861644.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:36Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812982d3f21ea3bbb17d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 737,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "35690",
            "path_kind": "directory",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
            "historical_root": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966/log_rank0.txt",
            "evidence_line": 619,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "11300",
                "14010",
                "16720",
                "19430",
                "22140",
                "24850",
                "27560",
                "30270",
                "3170",
                "32980",
                "35690",
                "460",
                "5880",
                "8590",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c250g250_GOOG_3861644.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c250g250_GOOG_3861644.out",
            "line": 17,
            "text": "  Using existing: /projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c250g250_GOOG_3861644.out",
            "line": 475,
            "text": "  Loading: scores_uncond_GOOG_phaseA-r000-ext-c250g250_integrated_3861644.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c250g250_GOOG_3701966.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 35690         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c250g250_GOOG_3701966.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c250g250_GOOG_3861644.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c250g250_GOOG_3861644.out",
            "line": 17,
            "text": "  Using existing: /projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c250g250_GOOG_3861644.out",
            "line": 475,
            "text": "  Loading: scores_uncond_GOOG_phaseA-r000-ext-c250g250_integrated_3861644.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r000-ext-c250g250/scores/scores_uncond_GOOG_phaseA-r000-ext-c250g250_integrated_3861644.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r000-ext-c250g250/scores/scores_uncond_GOOG_phaseA-r000-ext-c250g250_integrated_3861644.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.17476430272456195,
        "feature_points": {
          "ask_volume_touch": 0.1297056911918698,
          "ofi_up": 0.17631879478306023,
          "ask_volume": 0.12031132106279217,
          "limit_ask_order_depth": 0.1894987545389019,
          "ofi": 0.09920990789292276,
          "ofi_down": 0.0970003710024551,
          "limit_bid_order_depth": 0.28310034582096505,
          "orderbook_imbalance": 0.10200062676474139,
          "log_time_to_cancel": 0.2975131058901997,
          "ask_cancellation_depth": 0.15499603156116415,
          "spread": 0.267504015303606,
          "limit_bid_order_ticks": 0.2830709275234312,
          "bid_volume_touch": 0.12490750025809907,
          "limit_ask_order_ticks": 0.19279328331987608,
          "ask_cancellation_ticks": 0.1547661166642739,
          "bid_cancellation_depth": 0.2841789080799583,
          "bid_cancellation_ticks": 0.2829557047320109,
          "vol_per_min": 0.15417690047105878,
          "ofi_stay": 0.09899694613897131,
          "bid_volume": 0.10888289456138747,
          "log_inter_arrival_time": 0.06816220965405535
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r000-ext-c250g250/scores/scores_uncond_GOOG_phaseA-r000-ext-c250g250_integrated_3861644.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.174764,
            "report_record_id": "canonical-scores:37712c45-68fd-8129-82d3-f21ea3bbb17d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8129-8cd0-c051a537888f",
      "source_row_id": "37712c45-68fd-8129-8cd0-c051a537888f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81298cd0c051a537888f",
      "run": "24tok-s65073-L10sim",
      "stock": null,
      "model_size": null,
      "run_family": "24tok",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.216307,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.216307",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.179036,
      "reported_ks_context": 0.115683,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:44:03Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81298cd0c051a537888f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:44:03Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:44:03Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:44:03Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.216307,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8129-8cd0-c051a537888f",
        "class": "G20",
        "model_evaluation": "24tok-s65073-L10sim",
        "wws21_mean": 0.216307,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:44:03Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-s65073-L10sim_GOOG_3367853/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-s65073-L10sim_GOOG_3367853/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:44:03Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3302 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81298cd0c051a537888f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 738,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-s65073-L10sim_GOOG_3367853.out",
            "evidence_line": 8,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975 (step 65073)",
            "checkpoint_step": 65073,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-s65073-L10sim_GOOG_3367853.out",
                "evidence_line": 22,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-s65073-L10sim_GOOG_3367853\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 65073         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/.hf_indices_GOOG.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262799+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.121665+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-s65073-L10sim_GOOG_3367853.out",
            "line": 22,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-s65073-L10sim_GOOG_3367853\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 65073         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/.hf_indices_GOOG.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-s65073-L10sim_GOOG_3367853.out",
            "line": 14,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-s65073-L10sim_GOOG_3367853.out",
            "line": 19,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-s65073-L10sim_GOOG_3367853"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-s65073-L10sim_GOOG_3367853/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-s65073-L10sim_GOOG_3367853/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-s65073-L10sim_GOOG_3367853/data_real.tar",
            "selector": null,
            "source_value": 0.216307,
            "report_record_id": "canonical-scores:37712c45-68fd-8129-8cd0-c051a537888f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8129-9900-d067f06192ca",
      "source_row_id": "37712c45-68fd-8129-9900-d067f06192ca",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81299900d067f06192ca",
      "run": "R1-s4k-d256-step23840",
      "stock": null,
      "model_size": null,
      "run_family": "R1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.10051,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.10051",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.160673,
      "reported_ks_context": 0.096806,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:55Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81299900d067f06192ca",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:55Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:55Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:55Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.10051,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8129-9900-d067f06192ca",
        "class": "G20",
        "model_evaluation": "R1-s4k-d256-step23840",
        "wws21_mean": 0.10051,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:55Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step23840_GOOG_3740622/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step23840_GOOG_3740622/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:55Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81299900d067f06192ca",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 739,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133/23840",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step23840_GOOG_3740622.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133 (step 23840)",
            "checkpoint_step": 23840,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step23840_GOOG_3740622.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-step23840_GOOG_3740622\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 23840         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.146737+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step23840_GOOG_3740622.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-step23840_GOOG_3740622\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 23840         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step23840_GOOG_3740622.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step23840_GOOG_3740622.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-step23840_GOOG_3740622"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step23840_GOOG_3740622/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step23840_GOOG_3740622/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step23840_GOOG_3740622/data_real.tar",
            "selector": null,
            "source_value": 0.10051,
            "report_record_id": "canonical-scores:37712c45-68fd-8129-9900-d067f06192ca",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812a-9659-f2fb4823221a",
      "source_row_id": "37712c45-68fd-812a-9659-f2fb4823221a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812a9659f2fb4823221a",
      "run": "ctx4k-cond3500",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.154065,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.154065",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.123463,
      "reported_ks_context": 0.090329,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:02Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812a9659f2fb4823221a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:02Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:02Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:02Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.154065,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812a-9659-f2fb4823221a",
        "class": "G20",
        "model_evaluation": "ctx4k-cond3500",
        "wws21_mean": 0.154065,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:02Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond3500/scores/scores_uncond_GOOG_ctx4k-cond3500_integrated_2828765.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond3500_GOOG_2828765/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:02Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond3500_GOOG_2828765/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812a9659f2fb4823221a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 740,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond3500_GOOG_2828765/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 3500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond3500_GOOG_2828765.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond3500_GOOG_2828765\"         --n_cond_msgs 3500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond3500_GOOG_2828765.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond3500/scores/scores_uncond_GOOG_ctx4k-cond3500_integrated_2828765.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond3500/scores/scores_uncond_GOOG_ctx4k-cond3500_integrated_2828765.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15406469353945154,
        "feature_points": {
          "ask_cancellation_levels": 0.05103689412498784,
          "limit_bid_order_levels": 0.020303804839248075,
          "bid_cancellation_depth": 0.33035676048216567,
          "vol_per_min": 0.07415453212332793,
          "bid_cancellation_levels": 0.028073508533549358,
          "ask_volume_touch": 0.10626151468472902,
          "ofi": 0.036309321131879756,
          "limit_ask_order_levels": 0.055956725230794355,
          "spread": 0.2697480182410674,
          "orderbook_imbalance": 0.10154904533097757,
          "limit_bid_order_depth": 0.33730826356098165,
          "log_inter_arrival_time": 0.04809207678151099,
          "ask_cancellation_depth": 0.2950986940925686,
          "log_time_to_cancel": 0.3796449799256072,
          "ofi_up": 0.09188406381486622,
          "limit_ask_order_depth": 0.2560331358464886,
          "bid_volume": 0.31708084229964356,
          "ofi_stay": 0.036618461477965686,
          "ofi_down": 0.07305898792179598,
          "ask_volume": 0.2736763504104488,
          "bid_volume_touch": 0.053112583473878226
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond3500/scores/scores_uncond_GOOG_ctx4k-cond3500_integrated_2828765.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.154065,
            "report_record_id": "canonical-scores:37712c45-68fd-812a-9659-f2fb4823221a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812a-b73f-d1696ab6f97d",
      "source_row_id": "37712c45-68fd-812a-b73f-d1696ab6f97d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812ab73fd1696ab6f97d",
      "run": "pareto-55m-s68935",
      "stock": null,
      "model_size": "55M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.215924,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.215924",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.153741,
      "reported_ks_context": 0.116315,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:42Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812ab73fd1696ab6f97d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:42Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:42Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:42Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.215924,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812a-b73f-d1696ab6f97d",
        "class": "G20",
        "model_evaluation": "pareto-55m-s68935",
        "wws21_mean": 0.215924,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:42Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s68935_GOOG_3253391/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s68935_GOOG_3253391/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:42Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812ab73fd1696ab6f97d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 741,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/68935",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s68935_GOOG_3253391.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279 (step 68935)",
            "checkpoint_step": 68935,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s68935_GOOG_3253391.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s68935_GOOG_3253391\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 68935         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.136054+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s68935_GOOG_3253391.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s68935_GOOG_3253391\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 68935         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s68935_GOOG_3253391.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s68935_GOOG_3253391.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s68935_GOOG_3253391"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s68935_GOOG_3253391/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s68935_GOOG_3253391/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s68935_GOOG_3253391/data_real.tar",
            "selector": null,
            "source_value": 0.215924,
            "report_record_id": "canonical-scores:37712c45-68fd-812a-b73f-d1696ab6f97d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812c-84ee-e76bcaa569ed",
      "source_row_id": "37712c45-68fd-812c-84ee-e76bcaa569ed",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812c84eee76bcaa569ed",
      "run": "muon-s20496",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.200667,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.200667",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.146743,
      "reported_ks_context": 0.107851,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:48Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812c84eee76bcaa569ed",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:48Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:48Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:48Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.200667,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812c-84ee-e76bcaa569ed",
        "class": "G20",
        "model_evaluation": "muon-s20496",
        "wws21_mean": 0.200667,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:48Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s20496_GOOG_2868868/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s20496_GOOG_2868868/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:48Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812c84eee76bcaa569ed",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 742,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s20496_GOOG_2868868.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686 (step 20496)",
            "checkpoint_step": 20496,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s20496_GOOG_2868868.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s20496_GOOG_2868868\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 20496         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263052+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.148670+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s20496_GOOG_2868868.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s20496_GOOG_2868868\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 20496         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s20496_GOOG_2868868.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s20496_GOOG_2868868.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-s20496_GOOG_2868868"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s20496_GOOG_2868868/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s20496_GOOG_2868868/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s20496_GOOG_2868868/data_real.tar",
            "selector": null,
            "source_value": 0.200667,
            "report_record_id": "canonical-scores:37712c45-68fd-812c-84ee-e76bcaa569ed",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812c-b8d4-e8e73d793e6d",
      "source_row_id": "37712c45-68fd-812c-b8d4-e8e73d793e6d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812cb8d4e8e73d793e6d",
      "run": "nsa-o2b-prefill-v2",
      "stock": null,
      "model_size": null,
      "run_family": "nsa",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.744131,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.744131",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.683331,
      "reported_ks_context": 0.536548,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:52Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812cb8d4e8e73d793e6d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:52Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:52Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:52Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.744131,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812c-b8d4-e8e73d793e6d",
        "class": "G20",
        "model_evaluation": "nsa-o2b-prefill-v2",
        "wws21_mean": 0.744131,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:52Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-prefill-v2_GOOG_3252083/data_real",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-prefill-v2_GOOG_3252083/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:52Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812cb8d4e8e73d793e6d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 743,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922/14794",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-prefill-v2_GOOG_3252083.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922 (step 14794)",
            "checkpoint_step": 14794,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-prefill-v2_GOOG_3252083.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-prefill-v2_GOOG_3252083\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.148778+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-prefill-v2_GOOG_3252083.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-prefill-v2_GOOG_3252083\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-prefill-v2_GOOG_3252083.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-prefill-v2_GOOG_3252083.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/nsa-o2b-prefill-v2_GOOG_3252083"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-prefill-v2_GOOG_3252083/data_real",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-prefill-v2_GOOG_3252083/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-prefill-v2_GOOG_3252083/data_real",
            "selector": null,
            "source_value": 0.744131,
            "report_record_id": "canonical-scores:37712c45-68fd-812c-b8d4-e8e73d793e6d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812d-bcc4-c4f1c60d7e13",
      "source_row_id": "37712c45-68fd-812d-bcc4-c4f1c60d7e13",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812dbcc4c4f1c60d7e13",
      "run": "H2-55M-ctx4k-38k",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.165376,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.165376",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.130006,
      "reported_ks_context": 0.08997,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:22Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812dbcc4c4f1c60d7e13",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:22Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:22Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:22Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.165376,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812d-bcc4-c4f1c60d7e13",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-38k",
        "wws21_mean": 0.165376,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:22Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-38k/scores/scores_uncond_GOOG_H2-55M-ctx4k-38k_integrated_2572576.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-38k_GOOG_2572576/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:22Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-38k_GOOG_2572576/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812dbcc4c4f1c60d7e13",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 744,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "38335",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/38335",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-38k_GOOG_2572576/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/38335/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/38335",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/38335",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/38335",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-38k_GOOG_2572576.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-38k_GOOG_2572576\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 38335         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-38k_GOOG_2572576.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-38k/scores/scores_uncond_GOOG_H2-55M-ctx4k-38k_integrated_2572576.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-38k/scores/scores_uncond_GOOG_H2-55M-ctx4k-38k_integrated_2572576.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.16537585701991578,
        "feature_points": {
          "spread": 0.3724482703095987,
          "ask_cancellation_depth": 0.3375491143711026,
          "bid_cancellation_depth": 0.34310173727823995,
          "limit_ask_order_levels": 0.04116118629507604,
          "limit_bid_order_levels": 0.04173608369414754,
          "ask_cancellation_levels": 0.04623667096624397,
          "bid_cancellation_levels": 0.10012080109114463,
          "vol_per_min": 0.1541981340041942,
          "ofi": 0.11418199398340963,
          "ofi_up": 0.1192806124029926,
          "ofi_stay": 0.11300048927237875,
          "orderbook_imbalance": 0.11140931254100772,
          "ofi_down": 0.1322293060392083,
          "log_inter_arrival_time": 0.0636475029444499,
          "log_time_to_cancel": 0.35115643188806583,
          "ask_volume_touch": 0.08197906801614642,
          "bid_volume_touch": 0.06345402257323741,
          "ask_volume": 0.10314188309119723,
          "bid_volume": 0.1289556788851547,
          "limit_ask_order_depth": 0.3138588975081893,
          "limit_bid_order_depth": 0.34004580026304626
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-38k/scores/scores_uncond_GOOG_H2-55M-ctx4k-38k_integrated_2572576.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.165376,
            "report_record_id": "canonical-scores:37712c45-68fd-812d-bcc4-c4f1c60d7e13",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812d-bfec-fd01873dbf46",
      "source_row_id": "37712c45-68fd-812d-bfec-fd01873dbf46",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812dbfecfd01873dbf46",
      "run": "H2-55M-ctx4k-cond4000",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.133561,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.133561",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.115554,
      "reported_ks_context": 0.076537,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:20Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812dbfecfd01873dbf46",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:20Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:20Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:20Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.133561,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812d-bfec-fd01873dbf46",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond4000",
        "wws21_mean": 0.133561,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:20Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond4000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond4000_shard15_integrated_2632446.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond4000_GOOG_2632446/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:20Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond4000_GOOG_2632446/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812dbfecfd01873dbf46",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 745,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond4000_GOOG_2632446/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 4000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond4000_GOOG_2632446.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond4000_GOOG_2632446\"         --n_cond_msgs 4000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond4000_GOOG_2632446.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond4000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond4000_shard15_integrated_2632446.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond4000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond4000_integrated_2632446.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1335610409997162,
        "feature_points": {
          "spread": 0.1923876467494135,
          "ask_cancellation_depth": 0.32118488619367713,
          "bid_cancellation_depth": 0.32698539851784375,
          "limit_ask_order_levels": 0.01600399265614489,
          "limit_bid_order_levels": 0.02986064221574579,
          "ask_cancellation_levels": 0.033240869464174545,
          "bid_cancellation_levels": 0.033243587417625814,
          "vol_per_min": 0.06353703748194807,
          "ofi": 0.05115288350639255,
          "ofi_up": 0.0966988096016344,
          "ofi_stay": 0.051958609967502276,
          "orderbook_imbalance": 0.11021298608941346,
          "ofi_down": 0.07334269680762912,
          "log_inter_arrival_time": 0.043339805981618816,
          "log_time_to_cancel": 0.3805536692964992,
          "ask_volume_touch": 0.09714849855013834,
          "bid_volume_touch": 0.14595996723998922,
          "ask_volume": 0.11620956725264356,
          "bid_volume": 0.14500888249300392,
          "limit_ask_order_depth": 0.1684441405571033,
          "limit_bid_order_depth": 0.3083072829538984
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond4000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond4000_integrated_2632446.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.133561,
            "report_record_id": "canonical-scores:37712c45-68fd-812d-bfec-fd01873dbf46",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812e-8f5e-cece7cce3c90",
      "source_row_id": "37712c45-68fd-812e-8f5e-cece7cce3c90",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812e8f5ecece7cce3c90",
      "run": "nsa-o2b-v4-nobranch",
      "stock": null,
      "model_size": null,
      "run_family": "nsa",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.937769,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.937769",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.660617,
      "reported_ks_context": 0.580911,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:42Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812e8f5ecece7cce3c90",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:42Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:42Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:42Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.937769,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812e-8f5e-cece7cce3c90",
        "class": "G20",
        "model_evaluation": "nsa-o2b-v4-nobranch",
        "wws21_mean": 0.937769,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:42Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-v4-nobranch_GOOG_3234884/data_real",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-v4-nobranch_GOOG_3234884/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:42Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 128 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812e8f5ecece7cce3c90",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 746,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922/14794",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-v4-nobranch_GOOG_3234884.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922 (step 14794)",
            "checkpoint_step": 14794,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-v4-nobranch_GOOG_3234884.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-v4-nobranch_GOOG_3234884\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 128         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.148778+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-v4-nobranch_GOOG_3234884.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-v4-nobranch_GOOG_3234884\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 128         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-v4-nobranch_GOOG_3234884.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-v4-nobranch_GOOG_3234884.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/nsa-o2b-v4-nobranch_GOOG_3234884"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-v4-nobranch_GOOG_3234884/data_real",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-v4-nobranch_GOOG_3234884/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-v4-nobranch_GOOG_3234884/data_real",
            "selector": null,
            "source_value": 0.937769,
            "report_record_id": "canonical-scores:37712c45-68fd-812e-8f5e-cece7cce3c90",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812f-8da4-c53a9d8f0139",
      "source_row_id": "37712c45-68fd-812f-8da4-c53a9d8f0139",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812f8da4c53a9d8f0139",
      "run": "h200_23tok_step23k",
      "stock": null,
      "model_size": null,
      "run_family": "h200",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.231756,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.231756",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.189904,
      "reported_ks_context": 0.119551,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:02Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812f8da4c53a9d8f0139",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:02Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:02Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:02Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.231756,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812f-8da4-c53a9d8f0139",
        "class": "G20",
        "model_evaluation": "h200_23tok_step23k",
        "wws21_mean": 0.231756,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:02Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step23k/scores/scores_uncond_GOOG_h200_23tok_step23k_shard15_integrated_2570161.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566544/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:02Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566544/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812f8da4c53a9d8f0139",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 747,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "23310",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/23310",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566544/log_rank0.txt",
            "evidence_line": 628,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/23310/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/23310",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/23310",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "103505",
                "108205",
                "112915",
                "160000",
                "164710",
                "174140",
                "178853",
                "183557",
                "188270",
                "192987",
                "197705",
                "202424",
                "207134",
                "211843",
                "216558",
                "23310",
                "28025",
                "32740",
                "37455",
                "56320",
                "70465",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/23310",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step23k_GOOG_2570161.out",
            "line": 12,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step23k_GOOG_2570161.out",
            "line": 16,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566544"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step23k_GOOG_2570161.out",
            "line": 493,
            "text": "  - /projects/s5e/lob_pipeline/results_h200_23tok_step23k/scores/scores_uncond_GOOG_h200_23tok_step23k_shard15_integrated_2570161.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566544.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566544\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 23310         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566544.out",
            "line": 12,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step23k_GOOG_2570161.out",
            "line": 12,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step23k_GOOG_2570161.out",
            "line": 16,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566544"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step23k_GOOG_2570161.out",
            "line": 493,
            "text": "  - /projects/s5e/lob_pipeline/results_h200_23tok_step23k/scores/scores_uncond_GOOG_h200_23tok_step23k_shard15_integrated_2570161.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step23k/scores/scores_uncond_GOOG_h200_23tok_step23k_shard15_integrated_2570161.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step23k/scores/scores_uncond_GOOG_h200_23tok_step23k_integrated_2570161.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.23175622688048572,
        "feature_points": {
          "spread": 0.4228574483487797,
          "ask_cancellation_depth": 0.450635592914468,
          "bid_cancellation_depth": 0.472347188503442,
          "limit_ask_order_levels": 0.1174546570881266,
          "limit_bid_order_levels": 0.10914834105981358,
          "ask_cancellation_levels": 0.16848615565439748,
          "bid_cancellation_levels": 0.13567124666286237,
          "vol_per_min": 0.13371054418951422,
          "ofi": 0.11802831948540084,
          "ofi_up": 0.15313352338481845,
          "ofi_stay": 0.12045877340386857,
          "orderbook_imbalance": 0.11082441992647245,
          "ofi_down": 0.10881454275305076,
          "log_inter_arrival_time": 0.20927945307560433,
          "log_time_to_cancel": 0.6670623874738916,
          "ask_volume_touch": 0.16926273641660505,
          "bid_volume_touch": 0.10257804957957822,
          "ask_volume": 0.11102227410808888,
          "bid_volume": 0.07750528565099786,
          "limit_ask_order_depth": 0.4521070202383495,
          "limit_bid_order_depth": 0.4564928045720696
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step23k/scores/scores_uncond_GOOG_h200_23tok_step23k_integrated_2570161.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.231756,
            "report_record_id": "canonical-scores:37712c45-68fd-812f-8da4-c53a9d8f0139",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812f-8dc4-d9e776805f1d",
      "source_row_id": "37712c45-68fd-812f-8dc4-d9e776805f1d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812f8dc4d9e776805f1d",
      "run": "H1-120M-38k",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.202875,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.202875",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.153943,
      "reported_ks_context": 0.109364,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:17Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812f8dc4d9e776805f1d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:17Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:17Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:17Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.202875,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812f-8dc4-d9e776805f1d",
        "class": "G20",
        "model_evaluation": "H1-120M-38k",
        "wws21_mean": 0.202875,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:17Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-38k/scores/scores_uncond_GOOG_H1-120M-38k_integrated_2572575.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-38k_GOOG_2572575/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:17Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-38k_GOOG_2572575/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812f8dc4d9e776805f1d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 748,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "32645",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292/32645",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-38k_GOOG_2572575/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292/32645/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292/32645",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292/32645",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100669",
                "102860",
                "105056",
                "107254",
                "109459",
                "111653",
                "113844",
                "116029",
                "118227",
                "32645",
                "58990",
                "61190",
                "74350",
                "78745",
                "91900",
                "98480",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292/32645",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-38k_GOOG_2572575.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-120M-38k_GOOG_2572575\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 32645         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-38k_GOOG_2572575.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-38k/scores/scores_uncond_GOOG_H1-120M-38k_integrated_2572575.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-38k/scores/scores_uncond_GOOG_H1-120M-38k_integrated_2572575.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.20287535801023152,
        "feature_points": {
          "spread": 0.32730411511440194,
          "ask_cancellation_depth": 0.3567814885123314,
          "bid_cancellation_depth": 0.4043531250716711,
          "limit_ask_order_levels": 0.07209911278987807,
          "limit_bid_order_levels": 0.08789234706491816,
          "ask_cancellation_levels": 0.14219242004424704,
          "bid_cancellation_levels": 0.12312370436457895,
          "vol_per_min": 0.10584767863073072,
          "ofi": 0.08646210152847428,
          "ofi_up": 0.15875484182930902,
          "ofi_stay": 0.0852671038187779,
          "orderbook_imbalance": 0.09306101300988842,
          "ofi_down": 0.19294621169062984,
          "log_inter_arrival_time": 0.08772422641214704,
          "log_time_to_cancel": 0.45109417222526693,
          "ask_volume_touch": 0.18008261439637008,
          "bid_volume_touch": 0.22766819657829707,
          "ask_volume": 0.14831014793168032,
          "bid_volume": 0.17070531247263568,
          "limit_ask_order_depth": 0.36017444466848436,
          "limit_bid_order_depth": 0.39853814006014415
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-38k/scores/scores_uncond_GOOG_H1-120M-38k_integrated_2572575.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.202875,
            "report_record_id": "canonical-scores:37712c45-68fd-812f-8dc4-d9e776805f1d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-812f-a975-c1d774113f51",
      "source_row_id": "37712c45-68fd-812f-a975-c1d774113f51",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd812fa975c1d774113f51",
      "run": "phaseB-m3-14m-s1000-d64-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.206663,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.206663",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.207971,
      "reported_ks_context": 0.122296,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:49Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd812fa975c1d774113f51",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:49Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:49Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:49Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.206663,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-812f-a975-c1d774113f51",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s1000-d64-b-avg-c500g500",
        "wws21_mean": 0.206663,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:49Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:49Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd812fa975c1d774113f51",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 749,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934/log_rank0.txt",
            "evidence_line": 590,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703984_lxjyvwfw_3703984",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703984_lxjyvwfw_3703984\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703984_lxjyvwfw_3703984 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934/log_rank0.txt",
                "evidence_line": 592,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984",
            "availability_checked_at": "2026-09-12T06:13:15.372948+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703984_lxjyvwfw_3703984\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c500g500_GOOG_3713934/data_real.tar",
            "selector": null,
            "source_value": 0.206663,
            "report_record_id": "canonical-scores:37712c45-68fd-812f-a975-c1d774113f51",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8130-91af-c10439267311",
      "source_row_id": "37712c45-68fd-8130-91af-c10439267311",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd813091afc10439267311",
      "run": "H2-55M-ctx4k-cond1000",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.139077,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.139077",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.114082,
      "reported_ks_context": 0.077897,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:22Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd813091afc10439267311",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:22Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:22Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:22Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.139077,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8130-91af-c10439267311",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond1000",
        "wws21_mean": 0.139077,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:22Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond1000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond1000_shard4_integrated_2617323.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond1000_GOOG_2617323/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:22Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond1000_GOOG_2617323/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd813091afc10439267311",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 750,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond1000_GOOG_2617323/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond1000_GOOG_2617323.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond1000_GOOG_2617323\"         --n_cond_msgs 1000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond1000_GOOG_2617323.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond1000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond1000_shard4_integrated_2617323.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond1000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond1000_integrated_2617323.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1390769360179493,
        "feature_points": {
          "spread": 0.1900002711294171,
          "ask_cancellation_depth": 0.2773253354298811,
          "bid_cancellation_depth": 0.3305439630284626,
          "limit_ask_order_levels": 0.02978761011532744,
          "limit_bid_order_levels": 0.040318803942274384,
          "ask_cancellation_levels": 0.04124201307803672,
          "bid_cancellation_levels": 0.036139153509685454,
          "vol_per_min": 0.17816915830418387,
          "ofi": 0.03933193067925531,
          "ofi_up": 0.06659055892030952,
          "ofi_stay": 0.0392758672868136,
          "orderbook_imbalance": 0.1051332771015365,
          "ofi_down": 0.09263893412593496,
          "log_inter_arrival_time": 0.05012118741046143,
          "log_time_to_cancel": 0.3728710855162961,
          "ask_volume_touch": 0.1038245030553045,
          "bid_volume_touch": 0.09239307314186576,
          "ask_volume": 0.10721185104213028,
          "bid_volume": 0.09755884974556145,
          "limit_ask_order_depth": 0.34551680035539584,
          "limit_bid_order_depth": 0.2846214294588013
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond1000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond1000_integrated_2617323.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.139077,
            "report_record_id": "canonical-scores:37712c45-68fd-8130-91af-c10439267311",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8131-9f1a-ed326e8aed4d",
      "source_row_id": "37712c45-68fd-8131-9f1a-ed326e8aed4d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81319f1aed326e8aed4d",
      "run": "24tok-H1-s93224-L100sim-fix",
      "stock": null,
      "model_size": null,
      "run_family": "24tok",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.204122,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.204122",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.149505,
      "reported_ks_context": 0.113799,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:03Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81319f1aed326e8aed4d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:03Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:03Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:03Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.204122,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8131-9f1a-ed326e8aed4d",
        "class": "G20",
        "model_evaluation": "24tok-H1-s93224-L100sim-fix",
        "wws21_mean": 0.204122,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:03Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim-fix_GOOG_3234782/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim-fix_GOOG_3234782/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:03Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81319f1aed326e8aed4d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 751,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim-fix_GOOG_3234782.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279 (step 93224)",
            "checkpoint_step": 93224,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim-fix_GOOG_3234782.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-H1-s93224-L100sim-fix_GOOG_3234782\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 93224         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.136054+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim-fix_GOOG_3234782.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-H1-s93224-L100sim-fix_GOOG_3234782\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 93224         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim-fix_GOOG_3234782.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim-fix_GOOG_3234782.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-H1-s93224-L100sim-fix_GOOG_3234782"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim-fix_GOOG_3234782/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim-fix_GOOG_3234782/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim-fix_GOOG_3234782/data_real.tar",
            "selector": null,
            "source_value": 0.204122,
            "report_record_id": "canonical-scores:37712c45-68fd-8131-9f1a-ed326e8aed4d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8132-822a-c2cd76d71281",
      "source_row_id": "37712c45-68fd-8132-822a-c2cd76d71281",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8132822ac2cd76d71281",
      "run": "gdn-32head-s10800-c250g250-tokens",
      "stock": null,
      "model_size": null,
      "run_family": "gdn",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.426785,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.426785",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.740648,
      "reported_ks_context": 0.696638,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:42Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8132822ac2cd76d71281",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:42Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:42Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:42Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.426785,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8132-822a-c2cd76d71281",
        "class": "G20",
        "model_evaluation": "gdn-32head-s10800-c250g250-tokens",
        "wws21_mean": 0.426785,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:42Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c250g250-tokens_GOOG_3437864/data_real",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c250g250-tokens_GOOG_3437864/data_real;/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c250g250-tokens_GOOG_3437854/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:42Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8132822ac2cd76d71281",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 752,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273/10800",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c250g250-tokens_GOOG_3437864.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273 (step 10800)",
            "checkpoint_step": 10800,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c250g250-tokens_GOOG_3437864.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/gdn-32head-s10800-c250g250-tokens_GOOG_3437864\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 10800         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.151232+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c250g250-tokens_GOOG_3437864.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/gdn-32head-s10800-c250g250-tokens_GOOG_3437864\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 10800         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c250g250-tokens_GOOG_3437864.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c250g250-tokens_GOOG_3437864.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/gdn-32head-s10800-c250g250-tokens_GOOG_3437864"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c250g250-tokens_GOOG_3437864/data_real",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c250g250-tokens_GOOG_3437864/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c250g250-tokens_GOOG_3437864/data_real",
            "selector": null,
            "source_value": 0.426785,
            "report_record_id": "canonical-scores:37712c45-68fd-8132-822a-c2cd76d71281",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8132-937d-ca0ebbe39a01",
      "source_row_id": "37712c45-68fd-8132-937d-ca0ebbe39a01",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8132937dca0ebbe39a01",
      "run": "genlen-ctx2k-c500-g1000",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.226194,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.226194",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.185522,
      "reported_ks_context": 0.130569,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:40Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8132937dca0ebbe39a01",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:40Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:40Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:40Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.226194,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8132-937d-ca0ebbe39a01",
        "class": "G20",
        "model_evaluation": "genlen-ctx2k-c500-g1000",
        "wws21_mean": 0.226194,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:40Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1000_GOOG_3405781/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1000_GOOG_3405781/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:40Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8132937dca0ebbe39a01",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 753,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1000_GOOG_3405781.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1000_GOOG_3405781.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g1000_GOOG_3405781\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263020+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.139444+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 1000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1000_GOOG_3405781.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g1000_GOOG_3405781\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1000_GOOG_3405781.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1000_GOOG_3405781.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g1000_GOOG_3405781"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1000_GOOG_3405781/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1000_GOOG_3405781/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1000_GOOG_3405781/data_real.tar",
            "selector": null,
            "source_value": 0.226194,
            "report_record_id": "canonical-scores:37712c45-68fd-8132-937d-ca0ebbe39a01",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8132-b73a-ef27797fd6f9",
      "source_row_id": "37712c45-68fd-8132-b73a-ef27797fd6f9",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8132b73aef27797fd6f9",
      "run": "phaseA-r000-ext-c500g500",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.220829,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.220829",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.182706,
      "reported_ks_context": 0.123765,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:10Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8132b73aef27797fd6f9",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:10Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:10Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:10Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.220829,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8132-b73a-ef27797fd6f9",
        "class": "G20",
        "model_evaluation": "phaseA-r000-ext-c500g500",
        "wws21_mean": 0.220829,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:10Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r000-ext-c500g500/scores/scores_uncond_GOOG_phaseA-r000-ext-c500g500_integrated_3861646.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:10Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8132b73aef27797fd6f9",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 754,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "35690",
            "path_kind": "directory",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
            "historical_root": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967/log_rank0.txt",
            "evidence_line": 618,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "11300",
                "14010",
                "16720",
                "19430",
                "22140",
                "24850",
                "27560",
                "30270",
                "3170",
                "32980",
                "35690",
                "460",
                "5880",
                "8590",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c500g500_GOOG_3861646.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c500g500_GOOG_3861646.out",
            "line": 17,
            "text": "  Using existing: /projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c500g500_GOOG_3861646.out",
            "line": 683,
            "text": "  Loading: scores_uncond_GOOG_phaseA-r000-ext-c500g500_integrated_3861646.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c500g500_GOOG_3701967.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 35690         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c500g500_GOOG_3701967.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c500g500_GOOG_3861646.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c500g500_GOOG_3861646.out",
            "line": 17,
            "text": "  Using existing: /projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-ext-c500g500_GOOG_3861646.out",
            "line": 683,
            "text": "  Loading: scores_uncond_GOOG_phaseA-r000-ext-c500g500_integrated_3861646.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r000-ext-c500g500/scores/scores_uncond_GOOG_phaseA-r000-ext-c500g500_integrated_3861646.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r000-ext-c500g500/scores/scores_uncond_GOOG_phaseA-r000-ext-c500g500_integrated_3861646.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.22082879785031942,
        "feature_points": {
          "spread": 0.41932376521941156,
          "vol_per_min": 0.3027177826190041,
          "limit_bid_order_ticks": 0.3133661998702209,
          "ofi_down": 0.14328086401961773,
          "limit_bid_order_depth": 0.31377271336386386,
          "ofi_up": 0.15172956672171045,
          "limit_ask_order_depth": 0.2511703615550376,
          "bid_cancellation_depth": 0.2893976994174968,
          "ask_volume_touch": 0.13513095896587268,
          "ask_cancellation_depth": 0.24992476322175636,
          "bid_cancellation_ticks": 0.2886274561978773,
          "limit_ask_order_ticks": 0.25569251997613984,
          "ask_cancellation_ticks": 0.2497834679518069,
          "log_inter_arrival_time": 0.07332114562846524,
          "log_time_to_cancel": 0.3533220537724578,
          "orderbook_imbalance": 0.09067632855411427,
          "ofi": 0.10734471096002071,
          "ofi_stay": 0.10691374366686009,
          "bid_volume": 0.18687309537503072,
          "ask_volume": 0.22377554113326859,
          "bid_volume_touch": 0.13126001666667433
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r000-ext-c500g500/scores/scores_uncond_GOOG_phaseA-r000-ext-c500g500_integrated_3861646.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.220829,
            "report_record_id": "canonical-scores:37712c45-68fd-8132-b73a-ef27797fd6f9",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8133-bd85-de14c7f65551",
      "source_row_id": "37712c45-68fd-8133-bd85-de14c7f65551",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8133bd85de14c7f65551",
      "run": "H1-120M-cond2500",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.136099,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.136099",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.113625,
      "reported_ks_context": 0.076725,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:51Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8133bd85de14c7f65551",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:51Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:51Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:51Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.136099,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8133-bd85-de14c7f65551",
        "class": "G20",
        "model_evaluation": "H1-120M-cond2500",
        "wws21_mean": 0.136099,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:51Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond2500/scores/scores_uncond_GOOG_H1-120M-cond2500_shard18_integrated_2598226.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond2500_GOOG_2598226/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:51Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond2500_GOOG_2598226/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8133bd85de14c7f65551",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 755,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond2500_GOOG_2598226/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 2500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond2500_GOOG_2598226.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-120M-cond2500_GOOG_2598226\"         --n_cond_msgs 2500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 135458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond2500_GOOG_2598226.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond2500/scores/scores_uncond_GOOG_H1-120M-cond2500_shard18_integrated_2598226.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond2500/scores/scores_uncond_GOOG_H1-120M-cond2500_integrated_2598226.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.13609923183460876,
        "feature_points": {
          "spread": 0.20200729356567548,
          "ask_cancellation_depth": 0.20816909219839239,
          "bid_cancellation_depth": 0.29751828249045575,
          "limit_ask_order_levels": 0.09262840263453948,
          "limit_bid_order_levels": 0.07927113304670924,
          "ask_cancellation_levels": 0.10661912084533767,
          "bid_cancellation_levels": 0.07265215987667759,
          "vol_per_min": 0.06140672231210491,
          "ofi": 0.07251484329672418,
          "ofi_up": 0.10661199323830832,
          "ofi_stay": 0.07250599776298335,
          "orderbook_imbalance": 0.09967978297227487,
          "ofi_down": 0.09569074341474165,
          "log_inter_arrival_time": 0.06204208794024001,
          "log_time_to_cancel": 0.3535438052124508,
          "ask_volume_touch": 0.043075949310319925,
          "bid_volume_touch": 0.0861456800011495,
          "ask_volume": 0.07281321607380672,
          "bid_volume": 0.04644693226105094,
          "limit_ask_order_depth": 0.301287218530813,
          "limit_bid_order_depth": 0.32545341154202806
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond2500/scores/scores_uncond_GOOG_H1-120M-cond2500_integrated_2598226.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.136099,
            "report_record_id": "canonical-scores:37712c45-68fd-8133-bd85-de14c7f65551",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8134-965e-e08c8cabbfb2",
      "source_row_id": "37712c45-68fd-8134-965e-e08c8cabbfb2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8134965ee08c8cabbfb2",
      "run": "phaseB-m3-14m-s4000-d64-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.167659,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.167659",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.166051,
      "reported_ks_context": 0.097967,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:28Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8134965ee08c8cabbfb2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:28Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:28Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:28Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.167659,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8134-965e-e08c8cabbfb2",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s4000-d64-b-avg-c250g250",
        "wws21_mean": 0.167659,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:28Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:28Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8134965ee08c8cabbfb2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 756,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945/log_rank0.txt",
            "evidence_line": 574,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703799_69cc21ax_3703799",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703799_69cc21ax_3703799\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703799_69cc21ax_3703799 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945/log_rank0.txt",
                "evidence_line": 576,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799",
            "availability_checked_at": "2026-09-12T06:13:15.370080+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703799_69cc21ax_3703799/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703799_69cc21ax_3703799\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d64-b-avg-c250g250_GOOG_3713945/data_real.tar",
            "selector": null,
            "source_value": 0.167659,
            "report_record_id": "canonical-scores:37712c45-68fd-8134-965e-e08c8cabbfb2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8136-b08d-f46ff1699f19",
      "source_row_id": "37712c45-68fd-8136-b08d-f46ff1699f19",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8136b08df46ff1699f19",
      "run": "phaseA-base-a-s32120-c250g250",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.136122,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.136122",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.153899,
      "reported_ks_context": 0.098659,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:36Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8136b08df46ff1699f19",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:36Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:36Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:36Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.136122,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8136-b08d-f46ff1699f19",
        "class": "G20",
        "model_evaluation": "phaseA-base-a-s32120-c250g250",
        "wws21_mean": 0.136122,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:36Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c250g250_GOOG_3701963/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c250g250_GOOG_3701963/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:36Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8136b08df46ff1699f19",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 757,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344/32120",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c250g250_GOOG_3701963.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344 (step 32120)",
            "checkpoint_step": 32120,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c250g250_GOOG_3701963.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-base-a-s32120-c250g250_GOOG_3701963\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 32120         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.153453+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c250g250_GOOG_3701963.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-base-a-s32120-c250g250_GOOG_3701963\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 32120         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c250g250_GOOG_3701963.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c250g250_GOOG_3701963.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-base-a-s32120-c250g250_GOOG_3701963"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c250g250_GOOG_3701963/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c250g250_GOOG_3701963/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c250g250_GOOG_3701963/data_real.tar",
            "selector": null,
            "source_value": 0.136122,
            "report_record_id": "canonical-scores:37712c45-68fd-8136-b08d-f46ff1699f19",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8136-b491-fd27d24547b1",
      "source_row_id": "37712c45-68fd-8136-b491-fd27d24547b1",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8136b491fd27d24547b1",
      "run": "adamw-32n-s28434",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.162485,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.162485",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.126027,
      "reported_ks_context": 0.082859,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:18Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8136b491fd27d24547b1",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:18Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:18Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:18Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.162485,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8136-b491-fd27d24547b1",
        "class": "G20",
        "model_evaluation": "adamw-32n-s28434",
        "wws21_mean": 0.162485,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:18Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s28434_GOOG_2871387/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s28434_GOOG_2871387/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:18Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8136b491fd27d24547b1",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 758,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s28434_GOOG_2871387.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 28434)",
            "checkpoint_step": 28434,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s28434_GOOG_2871387.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s28434_GOOG_2871387\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 28434         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s28434_GOOG_2871387.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s28434_GOOG_2871387\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 28434         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s28434_GOOG_2871387.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s28434_GOOG_2871387.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s28434_GOOG_2871387"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s28434_GOOG_2871387/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s28434_GOOG_2871387/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s28434_GOOG_2871387/data_real.tar",
            "selector": null,
            "source_value": 0.162485,
            "report_record_id": "canonical-scores:37712c45-68fd-8136-b491-fd27d24547b1",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8137-8674-e523c45069fe",
      "source_row_id": "37712c45-68fd-8137-8674-e523c45069fe",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81378674e523c45069fe",
      "run": "soup_gdn_last5_gen500",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.18134,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.18134",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.131801,
      "reported_ks_context": 0.08985,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:00Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81378674e523c45069fe",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:00Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:00Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:00Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.18134,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8137-8674-e523c45069fe",
        "class": "G20",
        "model_evaluation": "soup_gdn_last5_gen500",
        "wws21_mean": 0.18134,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:00Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen500_GOOG_3150234/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen500_GOOG_3150234/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:00Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81378674e523c45069fe",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 759,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen500_GOOG_3150234.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen500_GOOG_3150234.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup_gdn_last5_gen500_GOOG_3150234\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262991+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134585+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen500_GOOG_3150234.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup_gdn_last5_gen500_GOOG_3150234\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen500_GOOG_3150234.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen500_GOOG_3150234.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup_gdn_last5_gen500_GOOG_3150234"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen500_GOOG_3150234/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen500_GOOG_3150234/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen500_GOOG_3150234/data_real.tar",
            "selector": null,
            "source_value": 0.18134,
            "report_record_id": "canonical-scores:37712c45-68fd-8137-8674-e523c45069fe",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8138-aa86-d9add587a730",
      "source_row_id": "37712c45-68fd-8138-aa86-d9add587a730",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8138aa86d9add587a730",
      "run": "genlen-ctx1k-c500-g500",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.221398,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.221398",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.179329,
      "reported_ks_context": 0.123237,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:55Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8138aa86d9add587a730",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:55Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:55Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:55Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.221398,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8138-aa86-d9add587a730",
        "class": "G20",
        "model_evaluation": "genlen-ctx1k-c500-g500",
        "wws21_mean": 0.221398,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:55Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g500_GOOG_3405777/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g500_GOOG_3405777/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:55Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8138aa86d9add587a730",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 760,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g500_GOOG_3405777.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g500_GOOG_3405777.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx1k-c500-g500_GOOG_3405777\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263039+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.145435+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g500_GOOG_3405777.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx1k-c500-g500_GOOG_3405777\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g500_GOOG_3405777.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g500_GOOG_3405777.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx1k-c500-g500_GOOG_3405777"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g500_GOOG_3405777/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g500_GOOG_3405777/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g500_GOOG_3405777/data_real.tar",
            "selector": null,
            "source_value": 0.221398,
            "report_record_id": "canonical-scores:37712c45-68fd-8138-aa86-d9add587a730",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-813a-9fa9-c233b4732fd6",
      "source_row_id": "37712c45-68fd-813a-9fa9-c233b4732fd6",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd813a9fa9c233b4732fd6",
      "run": "ticks-120m-rand",
      "stock": null,
      "model_size": "120M",
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.190929,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.190929",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.148528,
      "reported_ks_context": 0.096743,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:49Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd813a9fa9c233b4732fd6",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:49Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:49Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:49Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.190929,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-813a-9fa9-c233b4732fd6",
        "class": "G20",
        "model_evaluation": "ticks-120m-rand",
        "wws21_mean": 0.190929,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:49Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-120m-rand/scores/scores_uncond_GOOG_ticks-120m-rand_integrated_3278222.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:49Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd813a9fa9c233b4732fd6",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 761,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609/log_rank0.txt",
            "evidence_line": 634,
            "evidence_text": "[Checkpoint] Loading step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-120m-rand_GOOG_3278222.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-120m-rand_GOOG_3278222.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-120m-rand_GOOG_3278222.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-120m-rand_integrated_3278222.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-120m-rand_GOOG_2887609.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-120m-rand_GOOG_2887609.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-120m-rand_GOOG_3278222.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-120m-rand_GOOG_3278222.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-120m-rand_GOOG_3278222.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-120m-rand_integrated_3278222.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-120m-rand/scores/scores_uncond_GOOG_ticks-120m-rand_integrated_3278222.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-120m-rand/scores/scores_uncond_GOOG_ticks-120m-rand_integrated_3278222.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19092908218385607,
        "feature_points": {
          "ask_cancellation_depth": 0.3183279022625557,
          "bid_volume": 0.1255152976896818,
          "bid_volume_touch": 0.042998167654703934,
          "vol_per_min": 0.05672751789814602,
          "limit_bid_order_depth": 0.3456551461698399,
          "ofi_down": 0.050755516033257606,
          "ask_volume_touch": 0.03944415275829451,
          "spread": 0.19918459847910205,
          "bid_cancellation_depth": 0.3290333033438029,
          "ask_cancellation_ticks": 0.31768836126397065,
          "limit_ask_order_depth": 0.33102918516646707,
          "ofi": 0.08454706327762038,
          "bid_cancellation_ticks": 0.3282452575969955,
          "limit_ask_order_ticks": 0.33103726401097644,
          "ofi_stay": 0.08552442071957378,
          "limit_bid_order_ticks": 0.34563832517925636,
          "ofi_up": 0.04969733776006004,
          "ask_volume": 0.11180994569755244,
          "orderbook_imbalance": 0.10450510146905422,
          "log_time_to_cancel": 0.3493733260350006,
          "log_inter_arrival_time": 0.06277353539506518
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-120m-rand/scores/scores_uncond_GOOG_ticks-120m-rand_integrated_3278222.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.190929,
            "report_record_id": "canonical-scores:37712c45-68fd-813a-9fa9-c233b4732fd6",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-813c-a105-cca6a939b42d",
      "source_row_id": "37712c45-68fd-813c-a105-cca6a939b42d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd813ca105cca6a939b42d",
      "run": "H1-120M-cond1500",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.151134,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.151134",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.118505,
      "reported_ks_context": 0.085904,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd813ca105cca6a939b42d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.151134,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-813c-a105-cca6a939b42d",
        "class": "G20",
        "model_evaluation": "H1-120M-cond1500",
        "wws21_mean": 0.151134,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1500/scores/scores_uncond_GOOG_H1-120M-cond1500_shard8_integrated_2598227.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond1500_GOOG_2598227/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond1500_GOOG_2598227/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd813ca105cca6a939b42d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 762,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond1500_GOOG_2598227/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond1500_GOOG_2598227.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-120M-cond1500_GOOG_2598227\"         --n_cond_msgs 1500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 135458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond1500_GOOG_2598227.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1500/scores/scores_uncond_GOOG_H1-120M-cond1500_shard8_integrated_2598227.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1500/scores/scores_uncond_GOOG_H1-120M-cond1500_integrated_2598227.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15113419149724935,
        "feature_points": {
          "spread": 0.22154172123970278,
          "ask_cancellation_depth": 0.30998368883458516,
          "bid_cancellation_depth": 0.3166037590702749,
          "limit_ask_order_levels": 0.11017728292611016,
          "limit_bid_order_levels": 0.06733407972284489,
          "ask_cancellation_levels": 0.08784307285066159,
          "bid_cancellation_levels": 0.09305819537530634,
          "vol_per_min": 0.1265719225804809,
          "ofi": 0.11704455916290775,
          "ofi_up": 0.09566738091511223,
          "ofi_stay": 0.1160130096417282,
          "orderbook_imbalance": 0.09777292194184689,
          "ofi_down": 0.15637838178562546,
          "log_inter_arrival_time": 0.04668294726780128,
          "log_time_to_cancel": 0.3305426192139115,
          "ask_volume_touch": 0.08199086811296213,
          "bid_volume_touch": 0.05089712819220351,
          "ask_volume": 0.06981976837778024,
          "bid_volume": 0.07809596096840941,
          "limit_ask_order_depth": 0.27531267250885566,
          "limit_bid_order_depth": 0.3244860807531258
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1500/scores/scores_uncond_GOOG_H1-120M-cond1500_integrated_2598227.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.151134,
            "report_record_id": "canonical-scores:37712c45-68fd-813c-a105-cca6a939b42d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-813c-ab94-eec1e2447d4b",
      "source_row_id": "37712c45-68fd-813c-ab94-eec1e2447d4b",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd813cab94eec1e2447d4b",
      "run": "1tok-75m-lobbench",
      "stock": null,
      "model_size": "75M",
      "run_family": "1tok",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.26917,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.26917",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.706105,
      "reported_ks_context": 0.141971,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:42Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd813cab94eec1e2447d4b",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:42Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:42Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:42Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.26917,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-813c-ab94-eec1e2447d4b",
        "class": "G20",
        "model_evaluation": "1tok-75m-lobbench",
        "wws21_mean": 0.26917,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "estimated",
        "date_confidence": "estimated",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:42Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_1tok-75m-lobbench/scores/scores_uncond_GOOG_1tok-75m-lobbench_integrated_2857953.pkl",
        "input_path": "",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期及 class 为估计，置信度 high；Recorded full-January sampler, 1024 windows, c=500/g=500 and no fixed index override. All 208 matching command observations linked to established date sets cover 20 days; raw outputs for this run are missing.；评估完成日期未确认；来源原有日期 2026-06-06T21:34:42Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd813cab94eec1e2447d4b",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 763,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "177350",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_1tok-75m-lobbench_GOOG_2857953.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371 (step 177350)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371/177350",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "111420",
                "115585",
                "123900",
                "152760",
                "164890",
                "169041",
                "173188",
                "177350",
                "181488",
                "185639",
                "189804",
                "193967",
                "198110",
                "202269",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371/177350",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Source input_path was blank; checkpoint is from the integrated-job log matching the exact score filename job ID. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_1tok-75m-lobbench_GOOG_2857953.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/1tok-75m-lobbench_GOOG_2857953\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 32         --checkpoint_step 177350         --test_split 1.0 --token_mode 1tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_1tok-75m-lobbench_GOOG_2857953.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "incomplete_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_1tok-75m-lobbench/scores/scores_uncond_GOOG_1tok-75m-lobbench_integrated_2857953.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_1tok-75m-lobbench/scores/scores_uncond_GOOG_1tok-75m-lobbench_integrated_2857953.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 8,
        "nonfinite_features": [
          "ofi_down",
          "ask_cancellation_depth",
          "bid_cancellation_depth",
          "limit_ask_order_depth",
          "vol_per_min",
          "bid_cancellation_levels",
          "limit_bid_order_levels",
          "limit_ask_order_levels",
          "ofi_up",
          "limit_bid_order_depth",
          "ask_cancellation_levels",
          "log_time_to_cancel",
          "log_inter_arrival_time"
        ],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "reviewed_nonfinite_feature_counts",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_1tok-75m-lobbench/scores/scores_uncond_GOOG_1tok-75m-lobbench_integrated_2857953.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.26917,
            "report_record_id": "canonical-scores:37712c45-68fd-813c-ab94-eec1e2447d4b",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "The reviewed 21-feature score file contains nonfinite feature values. The reported source score is retained; it is not a verified complete finite 21-feature mean. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-813c-af76-e671853f4d3e",
      "source_row_id": "37712c45-68fd-813c-af76-e671853f4d3e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd813caf76e671853f4d3e",
      "run": "46m-s2k-d256-hdim128",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.16893,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.16893",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.207543,
      "reported_ks_context": 0.124749,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:03Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd813caf76e671853f4d3e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:03Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:03Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:03Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.16893,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-813c-af76-e671853f4d3e",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-hdim128",
        "wws21_mean": 0.16893,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:03Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-hdim128_GOOG_3873791/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-hdim128_GOOG_3873791/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:03Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd813caf76e671853f4d3e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 764,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865274_r3fnei5y_3865274/50970",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-hdim128_GOOG_3873791.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865274_r3fnei5y_3865274 (step 50970)",
            "checkpoint_step": 50970,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865274_r3fnei5y_3865274",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-hdim128_GOOG_3873791.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865274_r3fnei5y_3865274\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-hdim128_GOOG_3873791\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 50970         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.154754+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865274_r3fnei5y_3865274",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865274_r3fnei5y_3865274"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-hdim128_GOOG_3873791.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865274_r3fnei5y_3865274\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-hdim128_GOOG_3873791\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 50970         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-hdim128_GOOG_3873791.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-hdim128_GOOG_3873791.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-hdim128_GOOG_3873791"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-hdim128_GOOG_3873791/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-hdim128_GOOG_3873791/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-hdim128_GOOG_3873791/data_real.tar",
            "selector": null,
            "source_value": 0.16893,
            "report_record_id": "canonical-scores:37712c45-68fd-813c-af76-e671853f4d3e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8140-9157-d76506253d5f",
      "source_row_id": "37712c45-68fd-8140-9157-d76506253d5f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81409157d76506253d5f",
      "run": "gdn-j2722414-diag",
      "stock": null,
      "model_size": null,
      "run_family": "gdn",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.239473,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.239473",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.650189,
      "reported_ks_context": 0.477021,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-11T02:56:19Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2057",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-11T02:56:19Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-11T02:56:19Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-11T02:56:19Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.239473,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8140-9157-d76506253d5f",
        "class": "G20",
        "model_evaluation": "gdn-j2722414-diag",
        "wws21_mean": 0.239473,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-11T02:56:19Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_gdn-j2722414-diag/scores/scores_uncond_GOOG_gdn-j2722414-diag_integrated_2816334.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/gdn-j2722414-diag_GOOG_2816334/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-11T02:56:19Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/gdn-j2722414-diag_GOOG_2816334/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81409157d76506253d5f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 765,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "23356",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/gdn-j2722414-diag_GOOG_2816334/log_rank0.txt",
            "evidence_line": 757,
            "evidence_text": "[Checkpoint] Loading step=23356 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414/23356",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "10850",
                "12413",
                "13978",
                "15545",
                "17111",
                "18675",
                "20237",
                "21799",
                "23356",
                "9287",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414/23356",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-j2722414-diag_GOOG_2816334.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/gdn-j2722414-diag_GOOG_2816334\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 23356         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-j2722414-diag_GOOG_2816334.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "incomplete_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_gdn-j2722414-diag/scores/scores_uncond_GOOG_gdn-j2722414-diag_integrated_2816334.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_gdn-j2722414-diag/scores/scores_uncond_GOOG_gdn-j2722414-diag_integrated_2816334.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 19,
        "nonfinite_features": [
          "log_time_to_cancel",
          "log_inter_arrival_time"
        ],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "reviewed_nonfinite_feature_counts",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_gdn-j2722414-diag/scores/scores_uncond_GOOG_gdn-j2722414-diag_integrated_2816334.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.239473,
            "report_record_id": "canonical-scores:37712c45-68fd-8140-9157-d76506253d5f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "The reviewed 21-feature score file contains nonfinite feature values. The reported source score is retained; it is not a verified complete finite 21-feature mean. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8140-9caf-c29c37a3ac16",
      "source_row_id": "37712c45-68fd-8140-9caf-c29c37a3ac16",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81409cafc29c37a3ac16",
      "run": "46m-s2k-d256-adamw",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.083513,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.083513",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.16698,
      "reported_ks_context": 0.09414,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:44Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81409cafc29c37a3ac16",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:44Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:44Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:44Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.083513,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8140-9caf-c29c37a3ac16",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-adamw",
        "wws21_mean": 0.083513,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:44Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-adamw_GOOG_3873790/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-adamw_GOOG_3873790/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:44Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81409cafc29c37a3ac16",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 766,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865273_obqrdkeu_3865273/38990",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-adamw_GOOG_3873790.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865273_obqrdkeu_3865273 (step 38990)",
            "checkpoint_step": 38990,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865273_obqrdkeu_3865273",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-adamw_GOOG_3873790.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865273_obqrdkeu_3865273\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-adamw_GOOG_3873790\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 38990         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.156221+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865273_obqrdkeu_3865273",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865273_obqrdkeu_3865273"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-adamw_GOOG_3873790.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865273_obqrdkeu_3865273\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-adamw_GOOG_3873790\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 38990         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-adamw_GOOG_3873790.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-adamw_GOOG_3873790.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-adamw_GOOG_3873790"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-adamw_GOOG_3873790/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-adamw_GOOG_3873790/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-adamw_GOOG_3873790/data_real.tar",
            "selector": null,
            "source_value": 0.083513,
            "report_record_id": "canonical-scores:37712c45-68fd-8140-9caf-c29c37a3ac16",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8141-8f80-d51f2fddfec4",
      "source_row_id": "37712c45-68fd-8141-8f80-d51f2fddfec4",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81418f80d51f2fddfec4",
      "run": "phaseB-m3-14m-h032-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.14689,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.14689",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.225587,
      "reported_ks_context": 0.126179,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:54Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81418f80d51f2fddfec4",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:54Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:54Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:54Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.14689,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8141-8f80-d51f2fddfec4",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-h032-b-avg-c500g500",
        "wws21_mean": 0.14689,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:54Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:54Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81418f80d51f2fddfec4",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 767,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926/log_rank0.txt",
            "evidence_line": 611,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704414_ri9x2wsp_3704414",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704414_ri9x2wsp_3704414\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704414_ri9x2wsp_3704414 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926/log_rank0.txt",
                "evidence_line": 613,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414",
            "availability_checked_at": "2026-09-12T06:13:15.377226+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704414_ri9x2wsp_3704414\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c500g500_GOOG_3713926/data_real.tar",
            "selector": null,
            "source_value": 0.14689,
            "report_record_id": "canonical-scores:37712c45-68fd-8141-8f80-d51f2fddfec4",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8141-a306-dc35165e3dcd",
      "source_row_id": "37712c45-68fd-8141-a306-dc35165e3dcd",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8141a306dc35165e3dcd",
      "run": "ctx1k-cond1000",
      "stock": null,
      "model_size": null,
      "run_family": "ctx1k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.179401,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.179401",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.134651,
      "reported_ks_context": 0.098823,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:11Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8141a306dc35165e3dcd",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:11Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:11Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:11Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.179401,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8141-a306-dc35165e3dcd",
        "class": "G20",
        "model_evaluation": "ctx1k-cond1000",
        "wws21_mean": 0.179401,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:11Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx1k-cond1000/scores/scores_uncond_GOOG_ctx1k-cond1000_integrated_2826365.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx1k-cond1000_GOOG_2826365/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:11Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx1k-cond1000_GOOG_2826365/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8141a306dc35165e3dcd",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 768,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "93205",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx1k-cond1000_GOOG_2826365/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=93205 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230/93205",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "32395",
                "55635",
                "62605",
                "74222",
                "76545",
                "78868",
                "81191",
                "83493",
                "85811",
                "88136",
                "90459",
                "9150",
                "92785",
                "93205",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230/93205",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx1k-cond1000_GOOG_2826365.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx1k-cond1000_GOOG_2826365\"         --n_cond_msgs 1000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 93205         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx1k-cond1000_GOOG_2826365.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx1k-cond1000/scores/scores_uncond_GOOG_ctx1k-cond1000_integrated_2826365.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx1k-cond1000/scores/scores_uncond_GOOG_ctx1k-cond1000_integrated_2826365.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.17940077276680544,
        "feature_points": {
          "spread": 0.24546725723777643,
          "bid_cancellation_levels": 0.08379486659820368,
          "bid_volume_touch": 0.12851895198794522,
          "ask_volume_touch": 0.16403201870075299,
          "limit_bid_order_levels": 0.07621908029438351,
          "bid_cancellation_depth": 0.32498148954727385,
          "limit_bid_order_depth": 0.27154518803284444,
          "log_time_to_cancel": 0.336175757254158,
          "ofi": 0.09516350682576506,
          "orderbook_imbalance": 0.10774503461612303,
          "vol_per_min": 0.10381882020751847,
          "ofi_down": 0.17167957108427231,
          "ask_cancellation_depth": 0.26588534753163695,
          "ask_cancellation_levels": 0.10429399851239374,
          "limit_ask_order_levels": 0.06884839539507019,
          "ofi_up": 0.16216427745900305,
          "ask_volume": 0.2483253945984119,
          "bid_volume": 0.36760635615931325,
          "limit_ask_order_depth": 0.30799305597554005,
          "ofi_stay": 0.09265403508349856,
          "log_inter_arrival_time": 0.04050382500102951
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx1k-cond1000/scores/scores_uncond_GOOG_ctx1k-cond1000_integrated_2826365.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.179401,
            "report_record_id": "canonical-scores:37712c45-68fd-8141-a306-dc35165e3dcd",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8142-af1d-ef6a7a8cd15a",
      "source_row_id": "37712c45-68fd-8142-af1d-ef6a7a8cd15a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8142af1def6a7a8cd15a",
      "run": "phaseA-base-a-s32120-c500g500",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.153167,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.153167",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.163224,
      "reported_ks_context": 0.105724,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:32:06Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8142af1def6a7a8cd15a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:32:06Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:32:06Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:32:06Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.153167,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8142-af1d-ef6a7a8cd15a",
        "class": "G20",
        "model_evaluation": "phaseA-base-a-s32120-c500g500",
        "wws21_mean": 0.153167,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:32:06Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c500g500_GOOG_3701964/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c500g500_GOOG_3701964/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:32:06Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8142af1def6a7a8cd15a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 769,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344/32120",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c500g500_GOOG_3701964.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344 (step 32120)",
            "checkpoint_step": 32120,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c500g500_GOOG_3701964.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-base-a-s32120-c500g500_GOOG_3701964\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 32120         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.153453+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c500g500_GOOG_3701964.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696344_771f2ww8_3696344\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-base-a-s32120-c500g500_GOOG_3701964\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 32120         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c500g500_GOOG_3701964.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-base-a-s32120-c500g500_GOOG_3701964.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-base-a-s32120-c500g500_GOOG_3701964"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c500g500_GOOG_3701964/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c500g500_GOOG_3701964/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-base-a-s32120-c500g500_GOOG_3701964/data_real.tar",
            "selector": null,
            "source_value": 0.153167,
            "report_record_id": "canonical-scores:37712c45-68fd-8142-af1d-ef6a7a8cd15a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8142-bfa6-c0ee9e3922ec",
      "source_row_id": "37712c45-68fd-8142-bfa6-c0ee9e3922ec",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8142bfa6c0ee9e3922ec",
      "run": "46m-s2k-d256-lr7e3-step31170",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.15829,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.15829",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.17125,
      "reported_ks_context": 0.113752,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:59Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8142bfa6c0ee9e3922ec",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:59Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:59Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:59Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.15829,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8142-bfa6-c0ee9e3922ec",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-lr7e3-step31170",
        "wws21_mean": 0.15829,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:59Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-lr7e3-step31170_GOOG_3815021/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-lr7e3-step31170_GOOG_3815021/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:59Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8142bfa6c0ee9e3922ec",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 770,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3788987_8d14gr5r_3788987/31170",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-lr7e3-step31170_GOOG_3815021.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3788987_8d14gr5r_3788987 (step 31170)",
            "checkpoint_step": 31170,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3788987_8d14gr5r_3788987",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-lr7e3-step31170_GOOG_3815021.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3788987_8d14gr5r_3788987\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-lr7e3-step31170_GOOG_3815021\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 31170         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.157411+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3788987_8d14gr5r_3788987",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3788987_8d14gr5r_3788987"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-lr7e3-step31170_GOOG_3815021.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3788987_8d14gr5r_3788987\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-lr7e3-step31170_GOOG_3815021\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 31170         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-lr7e3-step31170_GOOG_3815021.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-lr7e3-step31170_GOOG_3815021.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-lr7e3-step31170_GOOG_3815021"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-lr7e3-step31170_GOOG_3815021/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-lr7e3-step31170_GOOG_3815021/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-lr7e3-step31170_GOOG_3815021/data_real.tar",
            "selector": null,
            "source_value": 0.15829,
            "report_record_id": "canonical-scores:37712c45-68fd-8142-bfa6-c0ee9e3922ec",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8143-b11f-e2952025f68f",
      "source_row_id": "37712c45-68fd-8143-b11f-e2952025f68f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8143b11fe2952025f68f",
      "run": "phaseA-r100-ext-c250g250",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.126195,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.126195",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.162333,
      "reported_ks_context": 0.099822,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:00Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8143b11fe2952025f68f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:00Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:00Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:00Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.126195,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8143-b11f-e2952025f68f",
        "class": "G20",
        "model_evaluation": "phaseA-r100-ext-c250g250",
        "wws21_mean": 0.126195,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:00Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r100-ext-c250g250/scores/scores_uncond_GOOG_phaseA-r100-ext-c250g250_integrated_3861643.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:00Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8143b11fe2952025f68f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 771,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "28980",
            "path_kind": "directory",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
            "historical_root": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972/log_rank0.txt",
            "evidence_line": 615,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "11380",
                "13580",
                "15780",
                "17980",
                "20180",
                "22380",
                "24580",
                "2580",
                "26780",
                "28980",
                "380",
                "4780",
                "6980",
                "9180",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c250g250_GOOG_3861643.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c250g250_GOOG_3861643.out",
            "line": 17,
            "text": "  Using existing: /projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c250g250_GOOG_3861643.out",
            "line": 475,
            "text": "  Loading: scores_uncond_GOOG_phaseA-r100-ext-c250g250_integrated_3861643.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c250g250_GOOG_3701972.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 28980         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c250g250_GOOG_3701972.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c250g250_GOOG_3861643.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c250g250_GOOG_3861643.out",
            "line": 17,
            "text": "  Using existing: /projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c250g250_GOOG_3701972"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c250g250_GOOG_3861643.out",
            "line": 475,
            "text": "  Loading: scores_uncond_GOOG_phaseA-r100-ext-c250g250_integrated_3861643.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r100-ext-c250g250/scores/scores_uncond_GOOG_phaseA-r100-ext-c250g250_integrated_3861643.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r100-ext-c250g250/scores/scores_uncond_GOOG_phaseA-r100-ext-c250g250_integrated_3861643.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.126195356719448,
        "feature_points": {
          "limit_bid_order_depth": 0.16998834367044016,
          "limit_ask_order_ticks": 0.08376334616653969,
          "limit_ask_order_depth": 0.08197915555052852,
          "bid_cancellation_depth": 0.1760164787712827,
          "ofi_up": 0.21873001606445167,
          "bid_volume_touch": 0.06264353125608874,
          "orderbook_imbalance": 0.0982352061982191,
          "spread": 0.21775852205062715,
          "ofi_down": 0.1405378555477314,
          "log_inter_arrival_time": 0.014740910488787753,
          "vol_per_min": 0.1472991089441637,
          "ask_volume_touch": 0.07975324919280884,
          "ask_cancellation_ticks": 0.09353215102719054,
          "ofi_stay": 0.1139263095133525,
          "bid_volume": 0.11172319423686376,
          "bid_cancellation_ticks": 0.17516243429867667,
          "ofi": 0.11288111454611657,
          "limit_bid_order_ticks": 0.170093838217318,
          "log_time_to_cancel": 0.20931879594832384,
          "ask_cancellation_depth": 0.09389637887183735,
          "ask_volume": 0.07812255054705962
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r100-ext-c250g250/scores/scores_uncond_GOOG_phaseA-r100-ext-c250g250_integrated_3861643.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.126195,
            "report_record_id": "canonical-scores:37712c45-68fd-8143-b11f-e2952025f68f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8144-8806-dd8cec958944",
      "source_row_id": "37712c45-68fd-8144-8806-dd8cec958944",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81448806dd8cec958944",
      "run": "H1-120M-cond2000",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.150442,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.150442",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.115512,
      "reported_ks_context": 0.083061,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:25Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81448806dd8cec958944",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:25Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:25Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:25Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.150442,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8144-8806-dd8cec958944",
        "class": "G20",
        "model_evaluation": "H1-120M-cond2000",
        "wws21_mean": 0.150442,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:25Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond2000/scores/scores_uncond_GOOG_H1-120M-cond2000_shard0_integrated_2602511.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond2000_GOOG_2602511/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:25Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond2000_GOOG_2602511/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81448806dd8cec958944",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 772,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond2000_GOOG_2602511/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 2000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond2000_GOOG_2602511.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-120M-cond2000_GOOG_2602511\"         --n_cond_msgs 2000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 135458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond2000_GOOG_2602511.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond2000/scores/scores_uncond_GOOG_H1-120M-cond2000_shard0_integrated_2602511.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond2000/scores/scores_uncond_GOOG_H1-120M-cond2000_integrated_2602511.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15044156274477447,
        "feature_points": {
          "spread": 0.22972666470099162,
          "ask_cancellation_depth": 0.33566913873575926,
          "bid_cancellation_depth": 0.3081362839399468,
          "limit_ask_order_levels": 0.10654165672562158,
          "limit_bid_order_levels": 0.07667190395665645,
          "ask_cancellation_levels": 0.10140533958219203,
          "bid_cancellation_levels": 0.07662260492516032,
          "vol_per_min": 0.11691329004078752,
          "ofi": 0.09008162544680429,
          "ofi_up": 0.11808286211930571,
          "ofi_stay": 0.08867207038868091,
          "orderbook_imbalance": 0.1018467441956643,
          "ofi_down": 0.11183383001571946,
          "log_inter_arrival_time": 0.056843737670579764,
          "log_time_to_cancel": 0.35675860526827746,
          "ask_volume_touch": 0.0829728751910006,
          "bid_volume_touch": 0.056818654877984265,
          "ask_volume": 0.06976888727181944,
          "bid_volume": 0.07390710950074009,
          "limit_ask_order_depth": 0.28604980632994986,
          "limit_bid_order_depth": 0.3139491267566218
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond2000/scores/scores_uncond_GOOG_H1-120M-cond2000_integrated_2602511.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.150442,
            "report_record_id": "canonical-scores:37712c45-68fd-8144-8806-dd8cec958944",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8144-8950-fd177ed3296c",
      "source_row_id": "37712c45-68fd-8144-8950-fd177ed3296c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81448950fd177ed3296c",
      "run": "pareto-120m-s23207",
      "stock": null,
      "model_size": "120M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.223009,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.223009",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.15694,
      "reported_ks_context": 0.120689,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:32:07Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81448950fd177ed3296c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:32:07Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:32:07Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:32:07Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.223009,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8144-8950-fd177ed3296c",
        "class": "G20",
        "model_evaluation": "pareto-120m-s23207",
        "wws21_mean": 0.223009,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:32:07Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s23207_GOOG_3253395/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s23207_GOOG_3253395/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:32:07Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81448950fd177ed3296c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 773,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260/23207",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s23207_GOOG_3253395.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260 (step 23207)",
            "checkpoint_step": 23207,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s23207_GOOG_3253395.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s23207_GOOG_3253395\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 23207         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.145456+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s23207_GOOG_3253395.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s23207_GOOG_3253395\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 23207         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s23207_GOOG_3253395.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s23207_GOOG_3253395.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s23207_GOOG_3253395"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s23207_GOOG_3253395/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s23207_GOOG_3253395/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s23207_GOOG_3253395/data_real.tar",
            "selector": null,
            "source_value": 0.223009,
            "report_record_id": "canonical-scores:37712c45-68fd-8144-8950-fd177ed3296c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8145-89de-c4482260d342",
      "source_row_id": "37712c45-68fd-8145-89de-c4482260d342",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd814589dec4482260d342",
      "run": "s4k-d256-c2000",
      "stock": null,
      "model_size": null,
      "run_family": "s4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.120992,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.120992",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.161468,
      "reported_ks_context": 0.09992,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:41Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd814589dec4482260d342",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:41Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:41Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:41Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.120992,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8145-89de-c4482260d342",
        "class": "G20",
        "model_evaluation": "s4k-d256-c2000",
        "wws21_mean": 0.120992,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:41Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-c2000_GOOG_3762574/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-c2000_GOOG_3762574/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:41Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd814589dec4482260d342",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 774,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133/23840",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-c2000_GOOG_3762574.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133 (step 23840)",
            "checkpoint_step": 23840,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-c2000_GOOG_3762574.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d256-c2000_GOOG_3762574\"         --n_cond_msgs 2000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 23840         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.146737+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 2000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-c2000_GOOG_3762574.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d256-c2000_GOOG_3762574\"         --n_cond_msgs 2000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 23840         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-c2000_GOOG_3762574.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-c2000_GOOG_3762574.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s4k-d256-c2000_GOOG_3762574"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-c2000_GOOG_3762574/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-c2000_GOOG_3762574/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-c2000_GOOG_3762574/data_real.tar",
            "selector": null,
            "source_value": 0.120992,
            "report_record_id": "canonical-scores:37712c45-68fd-8145-89de-c4482260d342",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8147-a1ad-e24775eb3305",
      "source_row_id": "37712c45-68fd-8147-a1ad-e24775eb3305",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8147a1ade24775eb3305",
      "run": "phaseB-m3-14m-s2000-d128-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.173187,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.173187",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.19255,
      "reported_ks_context": 0.12487,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:59Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8147a1ade24775eb3305",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:59Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:59Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:59Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.173187,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8147-a1ad-e24775eb3305",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s2000-d128-b-avg-c500g500",
        "wws21_mean": 0.173187,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:59Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:59Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8147a1ade24775eb3305",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 775,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936/log_rank0.txt",
            "evidence_line": 629,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704413_xgloeib5_3704413",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704413_xgloeib5_3704413\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704413_xgloeib5_3704413 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936/log_rank0.txt",
                "evidence_line": 631,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413",
            "availability_checked_at": "2026-09-12T06:13:15.376162+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704413_xgloeib5_3704413/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704413_xgloeib5_3704413\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d128-b-avg-c500g500_GOOG_3713936/data_real.tar",
            "selector": null,
            "source_value": 0.173187,
            "report_record_id": "canonical-scores:37712c45-68fd-8147-a1ad-e24775eb3305",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8148-9535-c95a8d861605",
      "source_row_id": "37712c45-68fd-8148-9535-c95a8d861605",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81489535c95a8d861605",
      "run": "phaseB-m3-14m-s2000-d64-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.165423,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.165423",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.215963,
      "reported_ks_context": 0.120628,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:45Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81489535c95a8d861605",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:45Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:45Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:45Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.165423,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8148-9535-c95a8d861605",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s2000-d64-b-avg-c250g250",
        "wws21_mean": 0.165423,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:45Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:45Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81489535c95a8d861605",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 776,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939/log_rank0.txt",
            "evidence_line": 584,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703796_haislv0g_3703796",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703796_haislv0g_3703796\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703796_haislv0g_3703796 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939/log_rank0.txt",
                "evidence_line": 586,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796",
            "availability_checked_at": "2026-09-12T06:13:15.368855+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703796_haislv0g_3703796\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c250g250_GOOG_3713939/data_real.tar",
            "selector": null,
            "source_value": 0.165423,
            "report_record_id": "canonical-scores:37712c45-68fd-8148-9535-c95a8d861605",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8148-b19f-c6383fdb6bdc",
      "source_row_id": "37712c45-68fd-8148-b19f-c6383fdb6bdc",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8148b19fc6383fdb6bdc",
      "run": "adamw-s9914",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.275647,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.275647",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.1928,
      "reported_ks_context": 0.149146,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:44Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8148b19fc6383fdb6bdc",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:44Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:44Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:44Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.275647,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8148-b19f-c6383fdb6bdc",
        "class": "G20",
        "model_evaluation": "adamw-s9914",
        "wws21_mean": 0.275647,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:44Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s9914_GOOG_2868482/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s9914_GOOG_2868482/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:44Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8148b19fc6383fdb6bdc",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 777,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s9914_GOOG_2868482.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747 (step 9914)",
            "checkpoint_step": 9914,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s9914_GOOG_2868482.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s9914_GOOG_2868482\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 9914         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263082+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158732+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s9914_GOOG_2868482.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s9914_GOOG_2868482\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 9914         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s9914_GOOG_2868482.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s9914_GOOG_2868482.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-s9914_GOOG_2868482"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s9914_GOOG_2868482/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s9914_GOOG_2868482/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s9914_GOOG_2868482/data_real.tar",
            "selector": null,
            "source_value": 0.275647,
            "report_record_id": "canonical-scores:37712c45-68fd-8148-b19f-c6383fdb6bdc",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8148-bae9-f2b756db0ebf",
      "source_row_id": "37712c45-68fd-8148-bae9-f2b756db0ebf",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8148bae9f2b756db0ebf",
      "run": "genlen-ctx4k-c500-g2000",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.266197,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.266197",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.203808,
      "reported_ks_context": 0.150169,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:43Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8148bae9f2b756db0ebf",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:43Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:43Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:43Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.266197,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8148-bae9-f2b756db0ebf",
        "class": "G20",
        "model_evaluation": "genlen-ctx4k-c500-g2000",
        "wws21_mean": 0.266197,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:43Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g2000_GOOG_3405787/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g2000_GOOG_3405787/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:43Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8148bae9f2b756db0ebf",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 778,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g2000_GOOG_3405787.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g2000_GOOG_3405787.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g2000_GOOG_3405787\"         --n_cond_msgs 500         --n_gen_msgs 2000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262982+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134305+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 2000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g2000_GOOG_3405787.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g2000_GOOG_3405787\"         --n_cond_msgs 500         --n_gen_msgs 2000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g2000_GOOG_3405787.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g2000_GOOG_3405787.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g2000_GOOG_3405787"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g2000_GOOG_3405787/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g2000_GOOG_3405787/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g2000_GOOG_3405787/data_real.tar",
            "selector": null,
            "source_value": 0.266197,
            "report_record_id": "canonical-scores:37712c45-68fd-8148-bae9-f2b756db0ebf",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-814b-aa6c-cae23e48098c",
      "source_row_id": "37712c45-68fd-814b-aa6c-cae23e48098c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd814baa6ccae23e48098c",
      "run": "phaseB-m3-14m-s500-d256-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.142929,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.142929",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.191218,
      "reported_ks_context": 0.108355,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:14Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd814baa6ccae23e48098c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:14Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:14Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:14Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.142929,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-814b-aa6c-cae23e48098c",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s500-d256-b-avg-c250g250",
        "wws21_mean": 0.142929,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:14Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:14Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd814baa6ccae23e48098c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 779,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949/log_rank0.txt",
            "evidence_line": 647,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704411_2qztmsnp_3704411",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704411_2qztmsnp_3704411\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704411_2qztmsnp_3704411 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949/log_rank0.txt",
                "evidence_line": 649,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411",
            "availability_checked_at": "2026-09-12T06:13:15.374336+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704411_2qztmsnp_3704411\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c250g250_GOOG_3713949/data_real.tar",
            "selector": null,
            "source_value": 0.142929,
            "report_record_id": "canonical-scores:37712c45-68fd-814b-aa6c-cae23e48098c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-814b-bd0e-d6cfeb582273",
      "source_row_id": "37712c45-68fd-814b-bd0e-d6cfeb582273",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd814bbd0ed6cfeb582273",
      "run": "nsa-o2b-reorder-64n",
      "stock": null,
      "model_size": null,
      "run_family": "nsa",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.907731,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.907731",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.683335,
      "reported_ks_context": 0.582728,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:00Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd814bbd0ed6cfeb582273",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:00Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:00Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:00Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.907731,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-814b-bd0e-d6cfeb582273",
        "class": "G20",
        "model_evaluation": "nsa-o2b-reorder-64n",
        "wws21_mean": 0.907731,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:00Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-reorder-64n_GOOG_3236711/data_real",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-reorder-64n_GOOG_3236711/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:00Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd814bbd0ed6cfeb582273",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 780,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922/14794",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-reorder-64n_GOOG_3236711.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922 (step 14794)",
            "checkpoint_step": 14794,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-reorder-64n_GOOG_3236711.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-reorder-64n_GOOG_3236711\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.148778+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-reorder-64n_GOOG_3236711.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-reorder-64n_GOOG_3236711\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-reorder-64n_GOOG_3236711.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-reorder-64n_GOOG_3236711.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/nsa-o2b-reorder-64n_GOOG_3236711"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-reorder-64n_GOOG_3236711/data_real",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-reorder-64n_GOOG_3236711/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-reorder-64n_GOOG_3236711/data_real",
            "selector": null,
            "source_value": 0.907731,
            "report_record_id": "canonical-scores:37712c45-68fd-814b-bd0e-d6cfeb582273",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-814b-bd34-e7e9452c9ec7",
      "source_row_id": "37712c45-68fd-814b-bd34-e7e9452c9ec7",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd814bbd34e7e9452c9ec7",
      "run": "j2633975-goog-L10",
      "stock": null,
      "model_size": null,
      "run_family": "j2633975",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.180298,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.180298",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.137605,
      "reported_ks_context": 0.101253,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-05T12:20:22Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2118",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-05T12:20:22Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-05T12:20:22Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-05T12:20:22Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.180298,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-814b-bd34-e7e9452c9ec7",
        "class": "G20",
        "model_evaluation": "j2633975-goog-L10",
        "wws21_mean": 0.180298,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-05T12:20:22Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L10_GOOG_2930532/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L10_GOOG_2930532/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-05T12:20:22Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd814bbd34e7e9452c9ec7",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 781,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L10_GOOG_2930532.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975 (step 65073)",
            "checkpoint_step": 65073,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L10_GOOG_2930532.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/j2633975-goog-L10_GOOG_2930532\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 65073         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262799+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.121665+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L10_GOOG_2930532.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/j2633975-goog-L10_GOOG_2930532\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 65073         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L10_GOOG_2930532.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L10_GOOG_2930532.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/j2633975-goog-L10_GOOG_2930532"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L10_GOOG_2930532/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L10_GOOG_2930532/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L10_GOOG_2930532/data_real.tar",
            "selector": null,
            "source_value": 0.180298,
            "report_record_id": "canonical-scores:37712c45-68fd-814b-bd34-e7e9452c9ec7",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-814c-91d6-d077f5dfb88b",
      "source_row_id": "37712c45-68fd-814c-91d6-d077f5dfb88b",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd814c91d6d077f5dfb88b",
      "run": "ctx4k-cond4500",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.135469,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.135469",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.123751,
      "reported_ks_context": 0.089199,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:46Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd814c91d6d077f5dfb88b",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:46Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:46Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:46Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.135469,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-814c-91d6-d077f5dfb88b",
        "class": "G20",
        "model_evaluation": "ctx4k-cond4500",
        "wws21_mean": 0.135469,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:46Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond4500/scores/scores_uncond_GOOG_ctx4k-cond4500_integrated_2828767.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond4500_GOOG_2828767/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:46Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond4500_GOOG_2828767/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd814c91d6d077f5dfb88b",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 782,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond4500_GOOG_2828767/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 4500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond4500_GOOG_2828767.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond4500_GOOG_2828767\"         --n_cond_msgs 4500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond4500_GOOG_2828767.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond4500/scores/scores_uncond_GOOG_ctx4k-cond4500_integrated_2828767.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond4500/scores/scores_uncond_GOOG_ctx4k-cond4500_integrated_2828767.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1354686285236362,
        "feature_points": {
          "limit_ask_order_levels": 0.05745150925158354,
          "limit_bid_order_depth": 0.3323911732129623,
          "ofi_up": 0.06687900719916018,
          "ofi_down": 0.0703448070571624,
          "ask_cancellation_levels": 0.046589837946217225,
          "spread": 0.24971402482916116,
          "ask_cancellation_depth": 0.2302281775758619,
          "vol_per_min": 0.1072439028622793,
          "limit_bid_order_levels": 0.0190923371529419,
          "ask_volume": 0.26664364802321316,
          "bid_cancellation_levels": 0.03301555521814308,
          "ofi": 0.04042045027388659,
          "bid_cancellation_depth": 0.3201617051312917,
          "orderbook_imbalance": 0.10719058412425823,
          "bid_volume_touch": 0.05307802654583753,
          "limit_ask_order_depth": 0.15593025193203247,
          "log_time_to_cancel": 0.3513663611208274,
          "ask_volume_touch": 0.12721740921883862,
          "bid_volume": 0.13204739774509686,
          "ofi_stay": 0.04152805640395562,
          "log_inter_arrival_time": 0.036306976171648624
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond4500/scores/scores_uncond_GOOG_ctx4k-cond4500_integrated_2828767.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.135469,
            "report_record_id": "canonical-scores:37712c45-68fd-814c-91d6-d077f5dfb88b",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-814c-9c1c-e0f8e3a64446",
      "source_row_id": "37712c45-68fd-814c-9c1c-e0f8e3a64446",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd814c9c1ce0f8e3a64446",
      "run": "phaseB-m3-14m-s2000-d256-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.093263,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.093263",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.149468,
      "reported_ks_context": 0.102307,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:41Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd814c9c1ce0f8e3a64446",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:41Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:41Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:41Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.093263,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-814c-9c1c-e0f8e3a64446",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s2000-d256-b-avg-c500g500",
        "wws21_mean": 0.093263,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:41Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:41Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd814c9c1ce0f8e3a64446",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 783,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938/log_rank0.txt",
            "evidence_line": 655,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704065_hpxxr5mb_3704065",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704065_hpxxr5mb_3704065\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704065_hpxxr5mb_3704065 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938/log_rank0.txt",
                "evidence_line": 657,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065",
            "availability_checked_at": "2026-09-12T06:13:15.374097+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704065_hpxxr5mb_3704065\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d256-b-avg-c500g500_GOOG_3713938/data_real.tar",
            "selector": null,
            "source_value": 0.093263,
            "report_record_id": "canonical-scores:37712c45-68fd-814c-9c1c-e0f8e3a64446",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-814d-b98a-febdd9510eb9",
      "source_row_id": "37712c45-68fd-814d-b98a-febdd9510eb9",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd814db98afebdd9510eb9",
      "run": "I3_120M_step136371",
      "stock": null,
      "model_size": "120M",
      "run_family": "I3",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.152842,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.152842",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.132815,
      "reported_ks_context": 0.09048,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:02Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd814db98afebdd9510eb9",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:02Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:02Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:02Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.152842,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-814d-b98a-febdd9510eb9",
        "class": "G20",
        "model_evaluation": "I3_120M_step136371",
        "wws21_mean": 0.152842,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:02Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3_120M_step136371/scores/scores_uncond_GOOG_I3_120M_step136371_shard3_integrated_2559910.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/I3_120M_step136371_GOOG_2559910/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:02Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/I3_120M_step136371_GOOG_2559910/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd814db98afebdd9510eb9",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 784,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "136371",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2554781_36bkcm4c_2554781/136371",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2554781_36bkcm4c_2554781",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/I3_120M_step136371_GOOG_2559910/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2554781_36bkcm4c_2554781/136371/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2554781_36bkcm4c_2554781/136371",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2554781_36bkcm4c_2554781/136371",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2554781_36bkcm4c_2554781/136371",
            "availability": "not_found",
            "path_scope": "checkpoint_step",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2554781_36bkcm4c_2554781/136371",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            },
            "relocation_candidates": [
              {
                "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_I3_effective_bsz_DELETING_91047/checkpoints/j2554781_36bkcm4c_2554781/136371",
                "original_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2554781_36bkcm4c_2554781/136371",
                "availability": "exists",
                "entries": [
                  "state",
                  "_CHECKPOINT_METADATA",
                  "metadata"
                ],
                "type": "directory",
                "path_kind": "directory",
                "note": "Possible relocated copy under _DELETING_91047; exact run and step directory exists, identity unconfirmed.",
                "identity_status": "unconfirmed_identity",
                "caveat": "Observed _DELETING_91047 directory. Exact run and step names match, but no authoritative rename record was inspected."
              }
            ]
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. An existing exact run/step candidate was found under an observed _DELETING_91047 directory. It remains a candidate because no authoritative rename record was inspected."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_I3_120M_step136371_GOOG_2559910.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2554781_36bkcm4c_2554781\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/I3_120M_step136371_GOOG_2559910\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 136371         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_I3_120M_step136371_GOOG_2559910.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3_120M_step136371/scores/scores_uncond_GOOG_I3_120M_step136371_shard3_integrated_2559910.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3_120M_step136371/scores/scores_uncond_GOOG_I3_120M_step136371_integrated_2559910.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15284225579486546,
        "feature_points": {
          "spread": 0.3051069967675028,
          "ask_cancellation_depth": 0.30522946175403043,
          "bid_cancellation_depth": 0.3286607135501268,
          "limit_ask_order_levels": 0.1006684657779374,
          "limit_bid_order_levels": 0.11302914864578015,
          "ask_cancellation_levels": 0.12455268976757142,
          "bid_cancellation_levels": 0.11262553189135828,
          "vol_per_min": 0.14474842708221686,
          "ofi": 0.0540511143579749,
          "ofi_up": 0.05841860683547913,
          "ofi_stay": 0.05447434685042028,
          "orderbook_imbalance": 0.10844814081129125,
          "ofi_down": 0.04923561885407769,
          "log_inter_arrival_time": 0.06787480214737009,
          "log_time_to_cancel": 0.4299009414111321,
          "ask_volume_touch": 0.06108951778482397,
          "bid_volume_touch": 0.056751671214321675,
          "ask_volume": 0.0655458939541188,
          "bid_volume": 0.04347046994394473,
          "limit_ask_order_depth": 0.3002674392802404,
          "limit_bid_order_depth": 0.32553737301045516
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3_120M_step136371/scores/scores_uncond_GOOG_I3_120M_step136371_integrated_2559910.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.152842,
            "report_record_id": "canonical-scores:37712c45-68fd-814d-b98a-febdd9510eb9",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-814e-a185-dfc634517c9c",
      "source_row_id": "37712c45-68fd-814e-a185-dfc634517c9c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd814ea185dfc634517c9c",
      "run": "h200_23tok_step193k",
      "stock": null,
      "model_size": null,
      "run_family": "h200",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.174024,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.174024",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.141805,
      "reported_ks_context": 0.088604,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:19Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd814ea185dfc634517c9c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:19Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:19Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:19Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.174024,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-814e-a185-dfc634517c9c",
        "class": "G20",
        "model_evaluation": "h200_23tok_step193k",
        "wws21_mean": 0.174024,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:19Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step193k/scores/scores_uncond_GOOG_h200_23tok_step193k_shard16_integrated_2570164.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:19Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd814ea185dfc634517c9c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 785,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "192987",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546/log_rank0.txt",
            "evidence_line": 628,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "103505",
                "108205",
                "112915",
                "160000",
                "164710",
                "174140",
                "178853",
                "183557",
                "188270",
                "192987",
                "197705",
                "202424",
                "207134",
                "211843",
                "216558",
                "23310",
                "28025",
                "32740",
                "37455",
                "56320",
                "70465",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step193k_GOOG_2570164.out",
            "line": 12,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step193k_GOOG_2570164.out",
            "line": 16,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step193k_GOOG_2570164.out",
            "line": 494,
            "text": "  - /projects/s5e/lob_pipeline/results_h200_23tok_step193k/scores/scores_uncond_GOOG_h200_23tok_step193k_shard16_integrated_2570164.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 192987         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 12,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step193k_GOOG_2570164.out",
            "line": 12,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step193k_GOOG_2570164.out",
            "line": 16,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step193k_GOOG_2570164.out",
            "line": 494,
            "text": "  - /projects/s5e/lob_pipeline/results_h200_23tok_step193k/scores/scores_uncond_GOOG_h200_23tok_step193k_shard16_integrated_2570164.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step193k/scores/scores_uncond_GOOG_h200_23tok_step193k_shard16_integrated_2570164.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step193k/scores/scores_uncond_GOOG_h200_23tok_step193k_integrated_2570164.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.17402401487036837,
        "feature_points": {
          "spread": 0.18274037723892755,
          "ask_cancellation_depth": 0.4117896995174461,
          "bid_cancellation_depth": 0.38592356077598766,
          "limit_ask_order_levels": 0.051314323861630096,
          "limit_bid_order_levels": 0.08006242538012454,
          "ask_cancellation_levels": 0.08316656162463579,
          "bid_cancellation_levels": 0.12382530221048828,
          "vol_per_min": 0.1439888681861637,
          "ofi": 0.08268800959700444,
          "ofi_up": 0.05968414163944544,
          "ofi_stay": 0.08504752286852803,
          "orderbook_imbalance": 0.10393075832255333,
          "ofi_down": 0.03626670029855909,
          "log_inter_arrival_time": 0.14127707815785293,
          "log_time_to_cancel": 0.5205188875183971,
          "ask_volume_touch": 0.10456351693460463,
          "bid_volume_touch": 0.09768640597002205,
          "ask_volume": 0.06313815333244661,
          "bid_volume": 0.08169185493527988,
          "limit_ask_order_depth": 0.4326174312848686,
          "limit_bid_order_depth": 0.38258273262276954
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step193k/scores/scores_uncond_GOOG_h200_23tok_step193k_integrated_2570164.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.174024,
            "report_record_id": "canonical-scores:37712c45-68fd-814e-a185-dfc634517c9c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8150-89c4-d672642ddc14",
      "source_row_id": "37712c45-68fd-8150-89c4-d672642ddc14",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815089c4d672642ddc14",
      "run": "J1-anchor100-61k-l100",
      "stock": null,
      "model_size": null,
      "run_family": "J1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.547768,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.547768",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.564327,
      "reported_ks_context": 0.436947,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:49Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815089c4d672642ddc14",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:49Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:49Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:49Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.547768,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8150-89c4-d672642ddc14",
        "class": "G20",
        "model_evaluation": "J1-anchor100-61k-l100",
        "wws21_mean": 0.547768,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:49Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor100-61k-l100/scores/scores_uncond_GOOG_J1-anchor100-61k-l100_integrated_2773510.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor100-61k-l100_GOOG_2773510/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:49Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor100-61k-l100_GOOG_2773510/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815089c4d672642ddc14",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 786,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "60980",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor100-61k-l100_GOOG_2773510/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=60980 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976/60980",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976/60980",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_J1-anchor100-61k-l100_GOOG_2773510.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633976_1ws5w6d4_2633976\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/J1-anchor100-61k-l100_GOOG_2773510\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 60980         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_J1-anchor100-61k-l100_GOOG_2773510.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "incomplete_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor100-61k-l100/scores/scores_uncond_GOOG_J1-anchor100-61k-l100_integrated_2773510.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor100-61k-l100/scores/scores_uncond_GOOG_J1-anchor100-61k-l100_integrated_2773510.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 20,
        "nonfinite_features": [
          "log_inter_arrival_time"
        ],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "reviewed_nonfinite_feature_counts",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor100-61k-l100/scores/scores_uncond_GOOG_J1-anchor100-61k-l100_integrated_2773510.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.547768,
            "report_record_id": "canonical-scores:37712c45-68fd-8150-89c4-d672642ddc14",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "The reviewed 21-feature score file contains nonfinite feature values. The reported source score is retained; it is not a verified complete finite 21-feature mean. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8150-ada0-e3058ea269a2",
      "source_row_id": "37712c45-68fd-8150-ada0-e3058ea269a2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8150ada0e3058ea269a2",
      "run": "genlen-ctx2k-c500-g500",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.19848,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.19848",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.163181,
      "reported_ks_context": 0.108073,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8150ada0e3058ea269a2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.19848,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8150-ada0-e3058ea269a2",
        "class": "G20",
        "model_evaluation": "genlen-ctx2k-c500-g500",
        "wws21_mean": 0.19848,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g500_GOOG_3405780/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g500_GOOG_3405780/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8150ada0e3058ea269a2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 787,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g500_GOOG_3405780.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g500_GOOG_3405780.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g500_GOOG_3405780\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263020+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.139444+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g500_GOOG_3405780.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g500_GOOG_3405780\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g500_GOOG_3405780.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g500_GOOG_3405780.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g500_GOOG_3405780"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g500_GOOG_3405780/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g500_GOOG_3405780/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g500_GOOG_3405780/data_real.tar",
            "selector": null,
            "source_value": 0.19848,
            "report_record_id": "canonical-scores:37712c45-68fd-8150-ada0-e3058ea269a2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8150-bedf-d3dc55b8e6a1",
      "source_row_id": "37712c45-68fd-8150-bedf-d3dc55b8e6a1",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8150bedfd3dc55b8e6a1",
      "run": "nsa-o2b-START-FIX",
      "stock": null,
      "model_size": null,
      "run_family": "nsa",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.935379,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.935379",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.687258,
      "reported_ks_context": 0.581425,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:32Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8150bedfd3dc55b8e6a1",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:32Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:32Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:32Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.935379,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8150-bedf-d3dc55b8e6a1",
        "class": "G20",
        "model_evaluation": "nsa-o2b-START-FIX",
        "wws21_mean": 0.935379,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:32Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-START-FIX_GOOG_3255916/data_real",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-START-FIX_GOOG_3255916/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:32Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8150bedfd3dc55b8e6a1",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 788,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922/14794",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-START-FIX_GOOG_3255916.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922 (step 14794)",
            "checkpoint_step": 14794,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-START-FIX_GOOG_3255916.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-START-FIX_GOOG_3255916\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.148778+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-START-FIX_GOOG_3255916.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-START-FIX_GOOG_3255916\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-START-FIX_GOOG_3255916.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-START-FIX_GOOG_3255916.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/nsa-o2b-START-FIX_GOOG_3255916"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-START-FIX_GOOG_3255916/data_real",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-START-FIX_GOOG_3255916/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-START-FIX_GOOG_3255916/data_real",
            "selector": null,
            "source_value": 0.935379,
            "report_record_id": "canonical-scores:37712c45-68fd-8150-bedf-d3dc55b8e6a1",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8151-b790-f8ab60a1726e",
      "source_row_id": "37712c45-68fd-8151-b790-f8ab60a1726e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8151b790f8ab60a1726e",
      "run": "s2k-d256-s42440",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.109499,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.109499",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.148455,
      "reported_ks_context": 0.09412,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:39Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8151b790f8ab60a1726e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:39Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:39Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:39Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.109499,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8151-b790-f8ab60a1726e",
        "class": "G20",
        "model_evaluation": "s2k-d256-s42440",
        "wws21_mean": 0.109499,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:39Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s42440_GOOG_3835483/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s42440_GOOG_3835483/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:39Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8151b790f8ab60a1726e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 789,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596/42440",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s42440_GOOG_3835483.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596 (step 42440)",
            "checkpoint_step": 42440,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s42440_GOOG_3835483.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s42440_GOOG_3835483\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 42440         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.121699+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s42440_GOOG_3835483.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s42440_GOOG_3835483\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 42440         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s42440_GOOG_3835483.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s42440_GOOG_3835483.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-d256-s42440_GOOG_3835483"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s42440_GOOG_3835483/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s42440_GOOG_3835483/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s42440_GOOG_3835483/data_real.tar",
            "selector": null,
            "source_value": 0.109499,
            "report_record_id": "canonical-scores:37712c45-68fd-8151-b790-f8ab60a1726e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8151-bc0d-d06920452cd5",
      "source_row_id": "37712c45-68fd-8151-bc0d-d06920452cd5",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8151bc0dd06920452cd5",
      "run": "s2k-d256-avg5-s51k-55k",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.112386,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.112386",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.165186,
      "reported_ks_context": 0.090311,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:31Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8151bc0dd06920452cd5",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:31Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:31Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:31Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.112386,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8151-bc0d-d06920452cd5",
        "class": "G20",
        "model_evaluation": "s2k-d256-avg5-s51k-55k",
        "wws21_mean": 0.112386,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:31Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-avg5-s51k-55k_GOOG_3835532/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-avg5-s51k-55k_GOOG_3835532/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:31Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8151bc0dd06920452cd5",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 790,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-avg5-s51k-55k_GOOG_3835532.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596 (step 999999)",
            "checkpoint_step": 999999,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-avg5-s51k-55k_GOOG_3835532.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-avg5-s51k-55k_GOOG_3835532\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:13:15.378633+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-avg5-s51k-55k_GOOG_3835532.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-avg5-s51k-55k_GOOG_3835532\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-avg5-s51k-55k_GOOG_3835532.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-avg5-s51k-55k_GOOG_3835532.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-d256-avg5-s51k-55k_GOOG_3835532"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-avg5-s51k-55k_GOOG_3835532/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-avg5-s51k-55k_GOOG_3835532/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-avg5-s51k-55k_GOOG_3835532/data_real.tar",
            "selector": null,
            "source_value": 0.112386,
            "report_record_id": "canonical-scores:37712c45-68fd-8151-bc0d-d06920452cd5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8151-bfc5-cd22e4cca9a0",
      "source_row_id": "37712c45-68fd-8151-bfc5-cd22e4cca9a0",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8151bfc5cd22e4cca9a0",
      "run": "ticks-muon-k1-rand",
      "stock": null,
      "model_size": null,
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.218655,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.218655",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.161693,
      "reported_ks_context": 0.10736,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:34Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8151bfc5cd22e4cca9a0",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:34Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:34Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:34Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.218655,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8151-bfc5-cd22e4cca9a0",
        "class": "G20",
        "model_evaluation": "ticks-muon-k1-rand",
        "wws21_mean": 0.218655,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:34Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-muon-k1-rand/scores/scores_uncond_GOOG_ticks-muon-k1-rand_integrated_3278223.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:34Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8151bfc5cd22e4cca9a0",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 791,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007/log_rank0.txt",
            "evidence_line": 634,
            "evidence_text": "[Checkpoint] Loading step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-muon-k1-rand_GOOG_3278223.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-muon-k1-rand_GOOG_3278223.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-muon-k1-rand_GOOG_3278223.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-muon-k1-rand_integrated_3278223.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-muon16n-K1-rand_GOOG_2896007.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-muon16n-K1-rand_GOOG_2896007.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-muon-k1-rand_GOOG_3278223.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-muon-k1-rand_GOOG_3278223.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-muon-k1-rand_GOOG_3278223.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-muon-k1-rand_integrated_3278223.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-muon-k1-rand/scores/scores_uncond_GOOG_ticks-muon-k1-rand_integrated_3278223.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-muon-k1-rand/scores/scores_uncond_GOOG_ticks-muon-k1-rand_integrated_3278223.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.21865456841268435,
        "feature_points": {
          "limit_bid_order_ticks": 0.35306602779933316,
          "ofi_up": 0.1122606909438269,
          "limit_ask_order_ticks": 0.34450018335526433,
          "limit_bid_order_depth": 0.35314190903805237,
          "bid_cancellation_ticks": 0.35082859748926465,
          "ask_volume": 0.13706759391133824,
          "limit_ask_order_depth": 0.3445822350889538,
          "ask_volume_touch": 0.08547661102753801,
          "ofi_down": 0.08213184641708887,
          "vol_per_min": 0.1318635861097088,
          "spread": 0.20059389500843378,
          "ask_cancellation_depth": 0.33524630502821817,
          "ask_cancellation_ticks": 0.3345738681869408,
          "bid_volume_touch": 0.07454267784464473,
          "orderbook_imbalance": 0.0995418324107143,
          "ofi": 0.08630453747542324,
          "bid_cancellation_depth": 0.35160980877901865,
          "bid_volume": 0.19455225963580097,
          "ofi_stay": 0.08568496225063772,
          "log_time_to_cancel": 0.4594228915094688,
          "log_inter_arrival_time": 0.0747536173567016
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-muon-k1-rand/scores/scores_uncond_GOOG_ticks-muon-k1-rand_integrated_3278223.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.218655,
            "report_record_id": "canonical-scores:37712c45-68fd-8151-bfc5-cd22e4cca9a0",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8152-b9d7-d23ae96cd168",
      "source_row_id": "37712c45-68fd-8152-b9d7-d23ae96cd168",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8152b9d7d23ae96cd168",
      "run": "s2k-cd-s101100",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.089568,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.089568",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.139055,
      "reported_ks_context": 0.091504,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:24Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8152b9d7d23ae96cd168",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:24Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:24Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:24Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.089568,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8152-b9d7-d23ae96cd168",
        "class": "G20",
        "model_evaluation": "s2k-cd-s101100",
        "wws21_mean": 0.089568,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:24Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s101100_GOOG_3864037/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s101100_GOOG_3864037/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:24Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8152b9d7d23ae96cd168",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 792,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/101100",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s101100_GOOG_3864037.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 101100)",
            "checkpoint_step": 101100,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s101100_GOOG_3864037.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s101100_GOOG_3864037\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 101100         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s101100_GOOG_3864037.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s101100_GOOG_3864037\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 101100         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s101100_GOOG_3864037.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s101100_GOOG_3864037.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s101100_GOOG_3864037"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s101100_GOOG_3864037/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s101100_GOOG_3864037/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s101100_GOOG_3864037/data_real.tar",
            "selector": null,
            "source_value": 0.089568,
            "report_record_id": "canonical-scores:37712c45-68fd-8152-b9d7-d23ae96cd168",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8153-8f66-c224275c917e",
      "source_row_id": "37712c45-68fd-8153-8f66-c224275c917e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81538f66c224275c917e",
      "run": "muon-s10190",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.228214,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.228214",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.172758,
      "reported_ks_context": 0.116217,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:09Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81538f66c224275c917e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:09Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:09Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:09Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.228214,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8153-8f66-c224275c917e",
        "class": "G20",
        "model_evaluation": "muon-s10190",
        "wws21_mean": 0.228214,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:09Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s10190_GOOG_2868867/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s10190_GOOG_2868867/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:09Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81538f66c224275c917e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 793,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s10190_GOOG_2868867.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686 (step 10190)",
            "checkpoint_step": 10190,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s10190_GOOG_2868867.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s10190_GOOG_2868867\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 10190         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263052+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.148670+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s10190_GOOG_2868867.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2696686_d8d071ln_2696686\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s10190_GOOG_2868867\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 10190         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s10190_GOOG_2868867.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s10190_GOOG_2868867.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-s10190_GOOG_2868867"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s10190_GOOG_2868867/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s10190_GOOG_2868867/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s10190_GOOG_2868867/data_real.tar",
            "selector": null,
            "source_value": 0.228214,
            "report_record_id": "canonical-scores:37712c45-68fd-8153-8f66-c224275c917e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8153-a031-d5d7a3c73649",
      "source_row_id": "37712c45-68fd-8153-a031-d5d7a3c73649",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8153a031d5d7a3c73649",
      "run": "pareto-360m-muon-s1255",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.298899,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.298899",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.284288,
      "reported_ks_context": 0.161078,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:17Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8153a031d5d7a3c73649",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:17Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:17Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:17Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.298899,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8153-a031-d5d7a3c73649",
        "class": "G20",
        "model_evaluation": "pareto-360m-muon-s1255",
        "wws21_mean": 0.298899,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:17Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s1255_GOOG_3253408/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s1255_GOOG_3253408/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:17Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8153a031d5d7a3c73649",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 794,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s1255_GOOG_3253408.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592 (step 1255)",
            "checkpoint_step": 1255,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s1255_GOOG_3253408.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s1255_GOOG_3253408\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 1255         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263086+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158865+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s1255_GOOG_3253408.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s1255_GOOG_3253408\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 1255         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s1255_GOOG_3253408.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s1255_GOOG_3253408.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s1255_GOOG_3253408"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s1255_GOOG_3253408/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s1255_GOOG_3253408/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s1255_GOOG_3253408/data_real.tar",
            "selector": null,
            "source_value": 0.298899,
            "report_record_id": "canonical-scores:37712c45-68fd-8153-a031-d5d7a3c73649",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8153-b3b7-e360c33c4760",
      "source_row_id": "37712c45-68fd-8153-b3b7-e360c33c4760",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8153b3b7e360c33c4760",
      "run": "pareto-360m-adamw32n-s34158",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.172704,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.172704",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.127337,
      "reported_ks_context": 0.088999,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:42Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8153b3b7e360c33c4760",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:42Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:42Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:42Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.172704,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8153-b3b7-e360c33c4760",
        "class": "G20",
        "model_evaluation": "pareto-360m-adamw32n-s34158",
        "wws21_mean": 0.172704,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:42Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s34158_GOOG_3253407/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s34158_GOOG_3253407/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:42Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8153b3b7e360c33c4760",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 795,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s34158_GOOG_3253407.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 34158)",
            "checkpoint_step": 34158,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s34158_GOOG_3253407.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw32n-s34158_GOOG_3253407\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 34158         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s34158_GOOG_3253407.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw32n-s34158_GOOG_3253407\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 34158         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s34158_GOOG_3253407.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s34158_GOOG_3253407.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw32n-s34158_GOOG_3253407"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s34158_GOOG_3253407/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s34158_GOOG_3253407/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s34158_GOOG_3253407/data_real.tar",
            "selector": null,
            "source_value": 0.172704,
            "report_record_id": "canonical-scores:37712c45-68fd-8153-b3b7-e360c33c4760",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8154-8d24-f0428db5db20",
      "source_row_id": "37712c45-68fd-8154-8d24-f0428db5db20",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81548d24f0428db5db20",
      "run": "24tok-H1-s93224-L100sim",
      "stock": null,
      "model_size": null,
      "run_family": "24tok",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.206031,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.206031",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.149495,
      "reported_ks_context": 0.114597,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:28Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81548d24f0428db5db20",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:28Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:28Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:28Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.206031,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8154-8d24-f0428db5db20",
        "class": "G20",
        "model_evaluation": "24tok-H1-s93224-L100sim",
        "wws21_mean": 0.206031,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:28Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim_GOOG_3234673/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim_GOOG_3234673/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:28Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81548d24f0428db5db20",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 796,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim_GOOG_3234673.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279 (step 93224)",
            "checkpoint_step": 93224,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim_GOOG_3234673.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-H1-s93224-L100sim_GOOG_3234673\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 93224         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.136054+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim_GOOG_3234673.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-H1-s93224-L100sim_GOOG_3234673\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 93224         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim_GOOG_3234673.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_24tok-H1-s93224-L100sim_GOOG_3234673.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/24tok-H1-s93224-L100sim_GOOG_3234673"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim_GOOG_3234673/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim_GOOG_3234673/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/24tok-H1-s93224-L100sim_GOOG_3234673/data_real.tar",
            "selector": null,
            "source_value": 0.206031,
            "report_record_id": "canonical-scores:37712c45-68fd-8154-8d24-f0428db5db20",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8154-a3b3-f4b7e515a3a2",
      "source_row_id": "37712c45-68fd-8154-a3b3-f4b7e515a3a2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8154a3b3f4b7e515a3a2",
      "run": "H1_75M_step93224",
      "stock": null,
      "model_size": "75M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.16269,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.16269",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.128806,
      "reported_ks_context": 0.089913,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:51Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8154a3b3f4b7e515a3a2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:51Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:51Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:51Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.16269,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8154-a3b3-f4b7e515a3a2",
        "class": "G20",
        "model_evaluation": "H1_75M_step93224",
        "wws21_mean": 0.16269,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:51Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_75M_step93224/scores/scores_uncond_GOOG_H1_75M_step93224_shard14_integrated_2559892.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1_75M_step93224_GOOG_2559892/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:51Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1_75M_step93224_GOOG_2559892/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8154a3b3f4b7e515a3a2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 797,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "93224",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1_75M_step93224_GOOG_2559892/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "68935",
                "75535",
                "77739",
                "79951",
                "82165",
                "84376",
                "86581",
                "88783",
                "90981",
                "93182",
                "93224",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279/93224",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1_75M_step93224_GOOG_2559892.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505279_85x1qq3d_2505279\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1_75M_step93224_GOOG_2559892\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 93224         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1_75M_step93224_GOOG_2559892.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_75M_step93224/scores/scores_uncond_GOOG_H1_75M_step93224_shard14_integrated_2559892.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_75M_step93224/scores/scores_uncond_GOOG_H1_75M_step93224_integrated_2559892.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.16268988847700974,
        "feature_points": {
          "spread": 0.2706105910910891,
          "ask_cancellation_depth": 0.3338231127115986,
          "bid_cancellation_depth": 0.34050027839960423,
          "limit_ask_order_levels": 0.044838937183136376,
          "limit_bid_order_levels": 0.07373516534450067,
          "ask_cancellation_levels": 0.07818862341720133,
          "bid_cancellation_levels": 0.09481521757013049,
          "vol_per_min": 0.13491277060606513,
          "ofi": 0.06829225473828812,
          "ofi_up": 0.13082789552477805,
          "ofi_stay": 0.06771712115200186,
          "orderbook_imbalance": 0.10890480474147475,
          "ofi_down": 0.0637419207717161,
          "log_inter_arrival_time": 0.04593593228717611,
          "log_time_to_cancel": 0.3866199301578144,
          "ask_volume_touch": 0.10785350201397546,
          "bid_volume_touch": 0.12824483245918852,
          "ask_volume": 0.10540581710401958,
          "bid_volume": 0.14187021668268646,
          "limit_ask_order_depth": 0.3407005959047842,
          "limit_bid_order_depth": 0.3489481381559748
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_75M_step93224/scores/scores_uncond_GOOG_H1_75M_step93224_integrated_2559892.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.16269,
            "report_record_id": "canonical-scores:37712c45-68fd-8154-a3b3-f4b7e515a3a2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8154-bb9d-e95ecaecd859",
      "source_row_id": "37712c45-68fd-8154-bb9d-e95ecaecd859",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8154bb9de95ecaecd859",
      "run": "soup-ctxt1k-rand",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.183476,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.183476",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.138862,
      "reported_ks_context": 0.104348,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:35Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8154bb9de95ecaecd859",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:35Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:35Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:35Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.183476,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8154-bb9d-e95ecaecd859",
        "class": "G20",
        "model_evaluation": "soup-ctxt1k-rand",
        "wws21_mean": 0.183476,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:35Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:35Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8154bb9de95ecaecd859",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 798,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt1k-rand_GOOG_2888139.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt1k-rand_GOOG_2888139.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263039+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.145435+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt1k-rand_GOOG_2888139.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt1k-rand_GOOG_2888139.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt1k-rand_GOOG_2888139.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt1k-rand_GOOG_2888139/data_real.tar",
            "selector": null,
            "source_value": 0.183476,
            "report_record_id": "canonical-scores:37712c45-68fd-8154-bb9d-e95ecaecd859",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8155-ab6d-dc11d0e79ac2",
      "source_row_id": "37712c45-68fd-8155-ab6d-dc11d0e79ac2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8155ab6ddc11d0e79ac2",
      "run": "rerun-mamba2-77M-soup10",
      "stock": null,
      "model_size": "77M",
      "run_family": "rerun",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.137953,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.137953",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.105795,
      "reported_ks_context": 0.075076,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:12Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8155ab6ddc11d0e79ac2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:12Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:12Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:12Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.137953,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8155-ab6d-dc11d0e79ac2",
        "class": "G20",
        "model_evaluation": "rerun-mamba2-77M-soup10",
        "wws21_mean": 0.137953,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:12Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:12Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8155ab6ddc11d0e79ac2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 799,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10/0",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.134324+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
            "selector": null,
            "source_value": 0.137953,
            "report_record_id": "canonical-scores:37712c45-68fd-8155-ab6d-dc11d0e79ac2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8156-a0de-fb3486364d0f",
      "source_row_id": "37712c45-68fd-8156-a0de-fb3486364d0f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8156a0defb3486364d0f",
      "run": "adamw-32n-s30342",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.141259,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.141259",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.112938,
      "reported_ks_context": 0.078123,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:11Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8156a0defb3486364d0f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:11Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:11Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:11Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.141259,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8156-a0de-fb3486364d0f",
        "class": "G20",
        "model_evaluation": "adamw-32n-s30342",
        "wws21_mean": 0.141259,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:11Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30342_GOOG_2871390/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30342_GOOG_2871390/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:11Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8156a0defb3486364d0f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 800,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30342_GOOG_2871390.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 30342)",
            "checkpoint_step": 30342,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30342_GOOG_2871390.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s30342_GOOG_2871390\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 30342         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30342_GOOG_2871390.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s30342_GOOG_2871390\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 30342         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30342_GOOG_2871390.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30342_GOOG_2871390.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s30342_GOOG_2871390"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30342_GOOG_2871390/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30342_GOOG_2871390/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30342_GOOG_2871390/data_real.tar",
            "selector": null,
            "source_value": 0.141259,
            "report_record_id": "canonical-scores:37712c45-68fd-8156-a0de-fb3486364d0f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8157-90e6-d97ac66c4820",
      "source_row_id": "37712c45-68fd-8157-90e6-d97ac66c4820",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815790e6d97ac66c4820",
      "run": "soup-muon16n-K1-rand",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.178734,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.178734",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.128236,
      "reported_ks_context": 0.093038,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:56Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815790e6d97ac66c4820",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:56Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:56Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:56Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.178734,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8157-90e6-d97ac66c4820",
        "class": "G20",
        "model_evaluation": "soup-muon16n-K1-rand",
        "wws21_mean": 0.178734,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:56Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:56Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815790e6d97ac66c4820",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 801,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-muon16n-K1-rand_GOOG_2896007.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-muon16n-K1-rand_GOOG_2896007.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263090+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158973+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-muon16n-K1-rand_GOOG_2896007.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_muon16n_K1_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-muon16n-K1-rand_GOOG_2896007.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-muon16n-K1-rand_GOOG_2896007.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-muon16n-K1-rand_GOOG_2896007/data_real.tar",
            "selector": null,
            "source_value": 0.178734,
            "report_record_id": "canonical-scores:37712c45-68fd-8157-90e6-d97ac66c4820",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8158-9016-f7134e60cd79",
      "source_row_id": "37712c45-68fd-8158-9016-f7134e60cd79",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81589016f7134e60cd79",
      "run": "H2-55M-ctx4k-cond3000",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.135361,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.135361",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.112684,
      "reported_ks_context": 0.07414,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:58Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81589016f7134e60cd79",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:58Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:58Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:58Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.135361,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8158-9016-f7134e60cd79",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond3000",
        "wws21_mean": 0.135361,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:58Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond3000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond3000_shard14_integrated_2617328.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond3000_GOOG_2617328/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:58Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond3000_GOOG_2617328/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81589016f7134e60cd79",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 802,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond3000_GOOG_2617328/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 3000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond3000_GOOG_2617328.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond3000_GOOG_2617328\"         --n_cond_msgs 3000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond3000_GOOG_2617328.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond3000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond3000_shard14_integrated_2617328.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond3000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond3000_integrated_2617328.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.13536099042406308,
        "feature_points": {
          "spread": 0.16604932548485735,
          "ask_cancellation_depth": 0.3234916287205417,
          "bid_cancellation_depth": 0.3137765878619723,
          "limit_ask_order_levels": 0.02083395305058487,
          "limit_bid_order_levels": 0.026732168607410164,
          "ask_cancellation_levels": 0.036854220766583796,
          "bid_cancellation_levels": 0.03405265263790678,
          "vol_per_min": 0.06694279906497141,
          "ofi": 0.03691857343546743,
          "ofi_up": 0.07471541613090749,
          "ofi_stay": 0.03718656413766508,
          "orderbook_imbalance": 0.09717856666542932,
          "ofi_down": 0.08884413304607645,
          "log_inter_arrival_time": 0.04684402557600809,
          "log_time_to_cancel": 0.3840730296550993,
          "ask_volume_touch": 0.09724855804356966,
          "bid_volume_touch": 0.14295724852080963,
          "ask_volume": 0.11497527553946377,
          "bid_volume": 0.10446965787311226,
          "limit_ask_order_depth": 0.2926386288390109,
          "limit_bid_order_depth": 0.3357977852478771
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond3000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond3000_integrated_2617328.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.135361,
            "report_record_id": "canonical-scores:37712c45-68fd-8158-9016-f7134e60cd79",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815a-be42-f92317e87558",
      "source_row_id": "37712c45-68fd-815a-be42-f92317e87558",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815abe42f92317e87558",
      "run": "46m-s4k-d512-lr7e3-step12580",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.173687,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.173687",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.252025,
      "reported_ks_context": 0.19201,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:40Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815abe42f92317e87558",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:40Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:40Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:40Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.173687,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815a-be42-f92317e87558",
        "class": "G20",
        "model_evaluation": "46m-s4k-d512-lr7e3-step12580",
        "wws21_mean": 0.173687,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:40Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-lr7e3-step12580_GOOG_3818100/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-lr7e3-step12580_GOOG_3818100/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:40Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815abe42f92317e87558",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 803,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3789403_mxhkks63_3789403/12580",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-lr7e3-step12580_GOOG_3818100.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3789403_mxhkks63_3789403 (step 12580)",
            "checkpoint_step": 12580,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3789403_mxhkks63_3789403",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-lr7e3-step12580_GOOG_3818100.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3789403_mxhkks63_3789403\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s4k-d512-lr7e3-step12580_GOOG_3818100\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 12580         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.158993+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3789403_mxhkks63_3789403",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3789403_mxhkks63_3789403"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-lr7e3-step12580_GOOG_3818100.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3789403_mxhkks63_3789403\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s4k-d512-lr7e3-step12580_GOOG_3818100\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 12580         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-lr7e3-step12580_GOOG_3818100.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-lr7e3-step12580_GOOG_3818100.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s4k-d512-lr7e3-step12580_GOOG_3818100"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-lr7e3-step12580_GOOG_3818100/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-lr7e3-step12580_GOOG_3818100/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-lr7e3-step12580_GOOG_3818100/data_real.tar",
            "selector": null,
            "source_value": 0.173687,
            "report_record_id": "canonical-scores:37712c45-68fd-815a-be42-f92317e87558",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815b-91c2-c55079623a77",
      "source_row_id": "37712c45-68fd-815b-91c2-c55079623a77",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815b91c2c55079623a77",
      "run": "soup-55m-rand",
      "stock": null,
      "model_size": "55M",
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.172861,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.172861",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.133778,
      "reported_ks_context": 0.095563,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:00Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815b91c2c55079623a77",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:00Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:00Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:00Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.172861,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815b-91c2-c55079623a77",
        "class": "G20",
        "model_evaluation": "soup-55m-rand",
        "wws21_mean": 0.172861,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:00Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:00Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815b91c2c55079623a77",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 804,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-55m-rand_GOOG_2887617.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-55m-rand_GOOG_2887617.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262945+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.125757+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-55m-rand_GOOG_2887617.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-55m-rand_GOOG_2887617.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-55m-rand_GOOG_2887617.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/data_real.tar",
            "selector": null,
            "source_value": 0.172861,
            "report_record_id": "canonical-scores:37712c45-68fd-815b-91c2-c55079623a77",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815c-9bc7-c52c6f492161",
      "source_row_id": "37712c45-68fd-815c-9bc7-c52c6f492161",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815c9bc7c52c6f492161",
      "run": "I3-120M-38k",
      "stock": null,
      "model_size": "120M",
      "run_family": "I3",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.198937,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.198937",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.148575,
      "reported_ks_context": 0.10915,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:20Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815c9bc7c52c6f492161",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:20Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:20Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:20Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.198937,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815c-9bc7-c52c6f492161",
        "class": "G20",
        "model_evaluation": "I3-120M-38k",
        "wws21_mean": 0.198937,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:20Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3-120M-38k/scores/scores_uncond_GOOG_I3-120M-38k_integrated_2572579.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/I3-120M-38k_GOOG_2572579/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:20Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/I3-120M-38k_GOOG_2572579/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815c9bc7c52c6f492161",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 805,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "38946",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/38946",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/I3-120M-38k_GOOG_2572579/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/38946/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/38946",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/38946",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/38946",
            "availability": "not_found",
            "path_scope": "checkpoint_step",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/38946",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            },
            "relocation_candidates": [
              {
                "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_I3_effective_bsz_DELETING_91047/checkpoints/j2518919_zqmh37u1_2518919/38946",
                "original_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/38946",
                "availability": "exists",
                "entries": [
                  "state",
                  "_CHECKPOINT_METADATA",
                  "metadata"
                ],
                "type": "directory",
                "path_kind": "directory",
                "note": "Possible relocated copy under _DELETING_91047; exact run and step directory exists, identity unconfirmed.",
                "identity_status": "unconfirmed_identity",
                "caveat": "Observed _DELETING_91047 directory. Exact run and step names match, but no authoritative rename record was inspected."
              }
            ]
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. An existing exact run/step candidate was found under an observed _DELETING_91047 directory. It remains a candidate because no authoritative rename record was inspected."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_I3-120M-38k_GOOG_2572579.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/I3-120M-38k_GOOG_2572579\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 38946         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_I3-120M-38k_GOOG_2572579.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3-120M-38k/scores/scores_uncond_GOOG_I3-120M-38k_integrated_2572579.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3-120M-38k/scores/scores_uncond_GOOG_I3-120M-38k_integrated_2572579.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19893705053248356,
        "feature_points": {
          "spread": 0.30502536429722676,
          "ask_cancellation_depth": 0.3268108590320992,
          "bid_cancellation_depth": 0.3294206259413819,
          "limit_ask_order_levels": 0.09622931517208128,
          "limit_bid_order_levels": 0.10022316252852213,
          "ask_cancellation_levels": 0.1466008055897686,
          "bid_cancellation_levels": 0.161459349721282,
          "vol_per_min": 0.2847503826031499,
          "ofi": 0.08036669080572428,
          "ofi_up": 0.13322367996030884,
          "ofi_stay": 0.07853513973789623,
          "orderbook_imbalance": 0.0804232848927034,
          "ofi_down": 0.14298317650230385,
          "log_inter_arrival_time": 0.13470888599824066,
          "log_time_to_cancel": 0.674284937879084,
          "ask_volume_touch": 0.12838730219697891,
          "bid_volume_touch": 0.11930955671669044,
          "ask_volume": 0.10278280515349189,
          "bid_volume": 0.07026295548699066,
          "limit_ask_order_depth": 0.33510342075024757,
          "limit_bid_order_depth": 0.3467863602159819
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3-120M-38k/scores/scores_uncond_GOOG_I3-120M-38k_integrated_2572579.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.198937,
            "report_record_id": "canonical-scores:37712c45-68fd-815c-9bc7-c52c6f492161",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815c-b95d-d0c4b6f50093",
      "source_row_id": "37712c45-68fd-815c-b95d-d0c4b6f50093",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815cb95dd0c4b6f50093",
      "run": "twilight-sound-77_jan2026",
      "stock": null,
      "model_size": null,
      "run_family": "twilight",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.368322,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.368322",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.25665,
      "reported_ks_context": 0.21585,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:15Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815cb95dd0c4b6f50093",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:15Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:15Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:15Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.368322,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815c-b95d-d0c4b6f50093",
        "class": "G20",
        "model_evaluation": "twilight-sound-77_jan2026",
        "wws21_mean": 0.368322,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:15Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_twilight-sound-77_jan2026/scores/scores_uncond_GOOG_twilight-sound-77_jan2026_shard2_integrated_2515481.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/twilight-sound-77_jan2026_GOOG_2515481/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:15Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/twilight-sound-77_jan2026_GOOG_2515481/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815cb95dd0c4b6f50093",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 806,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "23",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/lob_pipeline/best_checkpoints/twilight-sound-77_s42sujip/23",
            "historical_root": "/lus/lfs1aip2/projects/s5e/lob_pipeline/best_checkpoints/twilight-sound-77_s42sujip",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/twilight-sound-77_jan2026_GOOG_2515481/log_rank0.txt",
            "evidence_line": 630,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/lob_pipeline/best_checkpoints/twilight-sound-77_s42sujip/23/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/lob_pipeline/best_checkpoints/twilight-sound-77_s42sujip/23",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/lob_pipeline/best_checkpoints/twilight-sound-77_s42sujip/23",
              "availability": "permission_denied",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/lob_pipeline/best_checkpoints/twilight-sound-77_s42sujip/23",
            "availability": "not_found",
            "path_scope": "checkpoint_step",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/lob_pipeline/best_checkpoints/twilight-sound-77_s42sujip/23",
              "availability": "permission_denied",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. The mapped public candidate is permission denied; absence is not established there."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_twilight-sound-77_jan2026_GOOG_2515481.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/lob_pipeline/best_checkpoints/twilight-sound-77_s42sujip\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/twilight-sound-77_jan2026_GOOG_2515481\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 23         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_twilight-sound-77_jan2026_GOOG_2515481.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_twilight-sound-77_jan2026/scores/scores_uncond_GOOG_twilight-sound-77_jan2026_shard2_integrated_2515481.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_twilight-sound-77_jan2026/scores/scores_uncond_GOOG_twilight-sound-77_jan2026_integrated_2515481.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.36832247329840995,
        "feature_points": {
          "spread": 0.5729974208565843,
          "ask_cancellation_depth": 0.6604860728933304,
          "bid_cancellation_depth": 0.6841881757394255,
          "limit_ask_order_levels": 0.22721186968936416,
          "limit_bid_order_levels": 0.4084593340232503,
          "ask_cancellation_levels": 0.25721816271322484,
          "bid_cancellation_levels": 0.4040819248671434,
          "vol_per_min": 0.2136173542854156,
          "ofi": 0.2611482744245025,
          "ofi_up": 0.28551692826167485,
          "ofi_stay": 0.26460519852984526,
          "orderbook_imbalance": 0.1868203014149489,
          "ofi_down": 0.16517498905529152,
          "log_inter_arrival_time": 0.35548532600555516,
          "log_time_to_cancel": 0.10442506258994108,
          "ask_volume_touch": 0.1951508967056126,
          "bid_volume_touch": 0.15990423604928244,
          "ask_volume": 0.4823773297644694,
          "bid_volume": 0.5325278108000948,
          "limit_ask_order_depth": 0.6651688102983658,
          "limit_bid_order_depth": 0.6482064602992863
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_twilight-sound-77_jan2026/scores/scores_uncond_GOOG_twilight-sound-77_jan2026_integrated_2515481.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.368322,
            "report_record_id": "canonical-scores:37712c45-68fd-815c-b95d-d0c4b6f50093",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815d-8337-d5c4c247d02b",
      "source_row_id": "37712c45-68fd-815d-8337-d5c4c247d02b",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815d8337d5c4c247d02b",
      "run": "wsoup-adamw-32n-top5",
      "stock": null,
      "model_size": null,
      "run_family": "wsoup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.136054,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.136054",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.109087,
      "reported_ks_context": 0.070938,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:09Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815d8337d5c4c247d02b",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:09Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:09Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:09Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.136054,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815d-8337-d5c4c247d02b",
        "class": "G20",
        "model_evaluation": "wsoup-adamw-32n-top5",
        "wws21_mean": 0.136054,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:09Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-top5_GOOG_2871610/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-top5_GOOG_2871610/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:09Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815d8337d5c4c247d02b",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 807,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-top5_GOOG_2871610.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-top5_GOOG_2871610.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/wsoup-adamw-32n-top5_GOOG_2871610\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263098+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.160607+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-top5_GOOG_2871610.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/wsoup-adamw-32n-top5_GOOG_2871610\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-top5_GOOG_2871610.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-top5_GOOG_2871610.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/wsoup-adamw-32n-top5_GOOG_2871610"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-top5_GOOG_2871610/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-top5_GOOG_2871610/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-top5_GOOG_2871610/data_real.tar",
            "selector": null,
            "source_value": 0.136054,
            "report_record_id": "canonical-scores:37712c45-68fd-815d-8337-d5c4c247d02b",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815d-9102-cbb2dfa4be29",
      "source_row_id": "37712c45-68fd-815d-9102-cbb2dfa4be29",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815d9102cbb2dfa4be29",
      "run": "phaseA-r025-a-s34250-c500g500",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.113346,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.113346",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.194923,
      "reported_ks_context": 0.113767,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:09Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815d9102cbb2dfa4be29",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:09Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:09Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:09Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.113346,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815d-9102-cbb2dfa4be29",
        "class": "G20",
        "model_evaluation": "phaseA-r025-a-s34250-c500g500",
        "wws21_mean": 0.113346,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:09Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c500g500_GOOG_3701969/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c500g500_GOOG_3701969/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:09Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815d9102cbb2dfa4be29",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 808,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346/34250",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c500g500_GOOG_3701969.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346 (step 34250)",
            "checkpoint_step": 34250,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c500g500_GOOG_3701969.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r025-a-s34250-c500g500_GOOG_3701969\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 34250         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.160721+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c500g500_GOOG_3701969.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r025-a-s34250-c500g500_GOOG_3701969\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 34250         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c500g500_GOOG_3701969.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c500g500_GOOG_3701969.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-r025-a-s34250-c500g500_GOOG_3701969"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c500g500_GOOG_3701969/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c500g500_GOOG_3701969/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c500g500_GOOG_3701969/data_real.tar",
            "selector": null,
            "source_value": 0.113346,
            "report_record_id": "canonical-scores:37712c45-68fd-815d-9102-cbb2dfa4be29",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815d-b04b-e6dbdb5f7181",
      "source_row_id": "37712c45-68fd-815d-b04b-e6dbdb5f7181",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815db04be6dbdb5f7181",
      "run": "muon-s63568",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.157545,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.157545",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.120181,
      "reported_ks_context": 0.086849,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:04Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815db04be6dbdb5f7181",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:04Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:04Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:04Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.157545,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815d-b04b-e6dbdb5f7181",
        "class": "G20",
        "model_evaluation": "muon-s63568",
        "wws21_mean": 0.157545,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:04Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s63568_GOOG_2868872/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s63568_GOOG_2868872/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:04Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815db04be6dbdb5f7181",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 809,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s63568_GOOG_2868872.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925 (step 63568)",
            "checkpoint_step": 63568,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s63568_GOOG_2868872.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s63568_GOOG_2868872\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 63568         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263108+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.161994+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s63568_GOOG_2868872.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s63568_GOOG_2868872\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 63568         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s63568_GOOG_2868872.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s63568_GOOG_2868872.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-s63568_GOOG_2868872"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s63568_GOOG_2868872/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s63568_GOOG_2868872/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s63568_GOOG_2868872/data_real.tar",
            "selector": null,
            "source_value": 0.157545,
            "report_record_id": "canonical-scores:37712c45-68fd-815d-b04b-e6dbdb5f7181",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815e-9d4c-e926cd7a7c38",
      "source_row_id": "37712c45-68fd-815e-9d4c-e926cd7a7c38",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815e9d4ce926cd7a7c38",
      "run": "ticks-ctx500-rand",
      "stock": null,
      "model_size": null,
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.223094,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.223094",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.171242,
      "reported_ks_context": 0.115091,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:53Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815e9d4ce926cd7a7c38",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:53Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:53Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:53Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.223094,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815e-9d4c-e926cd7a7c38",
        "class": "G20",
        "model_evaluation": "ticks-ctx500-rand",
        "wws21_mean": 0.223094,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:53Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx500-rand/scores/scores_uncond_GOOG_ticks-ctx500-rand_integrated_3278227.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:53Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815e9d4ce926cd7a7c38",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 810,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/log_rank0.txt",
            "evidence_line": 634,
            "evidence_text": "[Checkpoint] Loading step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx500-rand_GOOG_3278227.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx500-rand_GOOG_3278227.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx500-rand_GOOG_3278227.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-ctx500-rand_integrated_3278227.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-55m-rand_GOOG_2887617.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-55m-rand_GOOG_2887617.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx500-rand_GOOG_3278227.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx500-rand_GOOG_3278227.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx500-rand_GOOG_3278227.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-ctx500-rand_integrated_3278227.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx500-rand/scores/scores_uncond_GOOG_ticks-ctx500-rand_integrated_3278227.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx500-rand/scores/scores_uncond_GOOG_ticks-ctx500-rand_integrated_3278227.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.22309377333810326,
        "feature_points": {
          "bid_cancellation_depth": 0.34410491504480195,
          "ask_cancellation_ticks": 0.33800261020929107,
          "limit_ask_order_depth": 0.3486699645177928,
          "limit_bid_order_depth": 0.357629773189992,
          "bid_volume_touch": 0.08925932777767114,
          "spread": 0.3042915771654641,
          "ofi_down": 0.0709777175037117,
          "vol_per_min": 0.09584228563871938,
          "limit_bid_order_ticks": 0.3575762781899478,
          "bid_cancellation_ticks": 0.3433577233457224,
          "limit_ask_order_ticks": 0.3487214667951836,
          "bid_volume": 0.2626913159755645,
          "ask_volume": 0.21780447215181653,
          "ofi_up": 0.08203170568861654,
          "log_time_to_cancel": 0.3997914588069087,
          "ask_cancellation_depth": 0.33860412757828656,
          "ofi_stay": 0.05238323024791695,
          "orderbook_imbalance": 0.1135731873204866,
          "ofi": 0.0511938899408996,
          "ask_volume_touch": 0.1117383134103259,
          "log_inter_arrival_time": 0.056723899601049066
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx500-rand/scores/scores_uncond_GOOG_ticks-ctx500-rand_integrated_3278227.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.223094,
            "report_record_id": "canonical-scores:37712c45-68fd-815e-9d4c-e926cd7a7c38",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815e-b591-c11d35ada435",
      "source_row_id": "37712c45-68fd-815e-b591-c11d35ada435",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815eb591c11d35ada435",
      "run": "wsoup-adamw-32n-L100eval",
      "stock": null,
      "model_size": null,
      "run_family": "wsoup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.244367,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.244367",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.153312,
      "reported_ks_context": 0.120153,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:34Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815eb591c11d35ada435",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:34Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:34Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:34Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.244367,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815e-b591-c11d35ada435",
        "class": "G20",
        "model_evaluation": "wsoup-adamw-32n-L100eval",
        "wws21_mean": 0.244367,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:34Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-L100eval_GOOG_2916960/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-L100eval_GOOG_2916960/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:34Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815eb591c11d35ada435",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 811,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-L100eval_GOOG_2916960.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-L100eval_GOOG_2916960.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/wsoup-adamw-32n-L100eval_GOOG_2916960\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263098+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.160607+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-L100eval_GOOG_2916960.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_32n_weighted\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/wsoup-adamw-32n-L100eval_GOOG_2916960\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-L100eval_GOOG_2916960.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_wsoup-adamw-32n-L100eval_GOOG_2916960.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/wsoup-adamw-32n-L100eval_GOOG_2916960"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-L100eval_GOOG_2916960/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-L100eval_GOOG_2916960/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/wsoup-adamw-32n-L100eval_GOOG_2916960/data_real.tar",
            "selector": null,
            "source_value": 0.244367,
            "report_record_id": "canonical-scores:37712c45-68fd-815e-b591-c11d35ada435",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815f-83aa-cca5792f9408",
      "source_row_id": "37712c45-68fd-815f-83aa-cca5792f9408",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815f83aacca5792f9408",
      "run": "phaseB-m3-14m-s2000-d64-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.175558,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.175558",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.229133,
      "reported_ks_context": 0.13356,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815f83aacca5792f9408",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.175558,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815f-83aa-cca5792f9408",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s2000-d64-b-avg-c500g500",
        "wws21_mean": 0.175558,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815f83aacca5792f9408",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 812,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940/log_rank0.txt",
            "evidence_line": 591,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703796_haislv0g_3703796",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703796_haislv0g_3703796\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703796_haislv0g_3703796 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940/log_rank0.txt",
                "evidence_line": 593,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796",
            "availability_checked_at": "2026-09-12T06:13:15.368855+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703796_haislv0g_3703796/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703796_haislv0g_3703796\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s2000-d64-b-avg-c500g500_GOOG_3713940/data_real.tar",
            "selector": null,
            "source_value": 0.175558,
            "report_record_id": "canonical-scores:37712c45-68fd-815f-83aa-cca5792f9408",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815f-8c46-d7a49d758162",
      "source_row_id": "37712c45-68fd-815f-8c46-d7a49d758162",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815f8c46d7a49d758162",
      "run": "G10-l10-bench",
      "stock": null,
      "model_size": null,
      "run_family": "G10",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 1.000411,
      "source_value_field": "Wasserstein",
      "source_value_text": "1.000411",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.840112,
      "reported_ks_context": 0.728485,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:40Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815f8c46d7a49d758162",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:40Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:40Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:40Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 1.000411,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815f-8c46-d7a49d758162",
        "class": "G20",
        "model_evaluation": "G10-l10-bench",
        "wws21_mean": 1.000411,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:40Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-bench_GOOG_2868245/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-bench_GOOG_2868245/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:40Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815f8c46d7a49d758162",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 813,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168/133168",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-bench_GOOG_2868245.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168 (step 133168)",
            "checkpoint_step": 133168,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-bench_GOOG_2868245.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l10-bench_GOOG_2868245\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 133168         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.162095+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-bench_GOOG_2868245.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l10-bench_GOOG_2868245\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 133168         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-bench_GOOG_2868245.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-bench_GOOG_2868245.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l10-bench_GOOG_2868245"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-bench_GOOG_2868245/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-bench_GOOG_2868245/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-bench_GOOG_2868245/data_real.tar",
            "selector": null,
            "source_value": 1.000411,
            "report_record_id": "canonical-scores:37712c45-68fd-815f-8c46-d7a49d758162",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-815f-8e90-f7abbd0c5131",
      "source_row_id": "37712c45-68fd-815f-8e90-f7abbd0c5131",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd815f8e90f7abbd0c5131",
      "run": "genlen-ctx500-c500-g500",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.233424,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.233424",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.177633,
      "reported_ks_context": 0.115226,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:38Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd815f8e90f7abbd0c5131",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:38Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:38Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:38Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.233424,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-815f-8e90-f7abbd0c5131",
        "class": "G20",
        "model_evaluation": "genlen-ctx500-c500-g500",
        "wws21_mean": 0.233424,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:38Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g500_GOOG_3405721/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g500_GOOG_3405721/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:38Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd815f8e90f7abbd0c5131",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 814,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g500_GOOG_3405721.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g500_GOOG_3405721.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx500-c500-g500_GOOG_3405721\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262945+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.125757+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g500_GOOG_3405721.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx500-c500-g500_GOOG_3405721\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g500_GOOG_3405721.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g500_GOOG_3405721.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx500-c500-g500_GOOG_3405721"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g500_GOOG_3405721/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g500_GOOG_3405721/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g500_GOOG_3405721/data_real.tar",
            "selector": null,
            "source_value": 0.233424,
            "report_record_id": "canonical-scores:37712c45-68fd-815f-8e90-f7abbd0c5131",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8161-9e6f-e0e93f7f6691",
      "source_row_id": "37712c45-68fd-8161-9e6f-e0e93f7f6691",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81619e6fe0e93f7f6691",
      "run": "H1-120M-cond3000",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.142865,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.142865",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.114498,
      "reported_ks_context": 0.079111,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:46Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81619e6fe0e93f7f6691",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:46Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:46Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:46Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.142865,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8161-9e6f-e0e93f7f6691",
        "class": "G20",
        "model_evaluation": "H1-120M-cond3000",
        "wws21_mean": 0.142865,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:46Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond3000/scores/scores_uncond_GOOG_H1-120M-cond3000_integrated_2598225.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond3000_GOOG_2598225/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:46Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond3000_GOOG_2598225/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81619e6fe0e93f7f6691",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 815,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond3000_GOOG_2598225/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 3000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond3000_GOOG_2598225.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-120M-cond3000_GOOG_2598225\"         --n_cond_msgs 3000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 135458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond3000_GOOG_2598225.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond3000/scores/scores_uncond_GOOG_H1-120M-cond3000_integrated_2598225.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond3000/scores/scores_uncond_GOOG_H1-120M-cond3000_integrated_2598225.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14286452010676232,
        "feature_points": {
          "spread": 0.19924511221780014,
          "ask_cancellation_depth": 0.30456531448526847,
          "bid_cancellation_depth": 0.3008954207747046,
          "limit_ask_order_levels": 0.11650479389992514,
          "limit_bid_order_levels": 0.07153570255682173,
          "ask_cancellation_levels": 0.08348960856527649,
          "bid_cancellation_levels": 0.09215905797349767,
          "vol_per_min": 0.10807135031039261,
          "ofi": 0.09155327965155738,
          "ofi_up": 0.07334889566297245,
          "ofi_stay": 0.09215680848085417,
          "orderbook_imbalance": 0.09582861794104806,
          "ofi_down": 0.08941281688980433,
          "log_inter_arrival_time": 0.058850181132002494,
          "log_time_to_cancel": 0.36620609244345625,
          "ask_volume_touch": 0.06550470175117337,
          "bid_volume_touch": 0.06656691036302777,
          "ask_volume": 0.06157016604104226,
          "bid_volume": 0.05472816372069953,
          "limit_ask_order_depth": 0.2744056506706824,
          "limit_bid_order_depth": 0.3335562767100016
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond3000/scores/scores_uncond_GOOG_H1-120M-cond3000_integrated_2598225.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.142865,
            "report_record_id": "canonical-scores:37712c45-68fd-8161-9e6f-e0e93f7f6691",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8161-a112-e8a6cf62d400",
      "source_row_id": "37712c45-68fd-8161-a112-e8a6cf62d400",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8161a112e8a6cf62d400",
      "run": "s2k-cd-s104660",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.092149,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.092149",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.142256,
      "reported_ks_context": 0.092068,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:49Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8161a112e8a6cf62d400",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:49Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:49Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:49Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.092149,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8161-a112-e8a6cf62d400",
        "class": "G20",
        "model_evaluation": "s2k-cd-s104660",
        "wws21_mean": 0.092149,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:49Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s104660_GOOG_3864038/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s104660_GOOG_3864038/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:49Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8161a112e8a6cf62d400",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 816,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/104660",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s104660_GOOG_3864038.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 104660)",
            "checkpoint_step": 104660,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s104660_GOOG_3864038.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s104660_GOOG_3864038\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 104660         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s104660_GOOG_3864038.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s104660_GOOG_3864038\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 104660         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s104660_GOOG_3864038.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s104660_GOOG_3864038.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s104660_GOOG_3864038"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s104660_GOOG_3864038/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s104660_GOOG_3864038/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s104660_GOOG_3864038/data_real.tar",
            "selector": null,
            "source_value": 0.092149,
            "report_record_id": "canonical-scores:37712c45-68fd-8161-a112-e8a6cf62d400",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8161-a699-d40e209add2d",
      "source_row_id": "37712c45-68fd-8161-a699-d40e209add2d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8161a699d40e209add2d",
      "run": "s2k-cooldown-s92200",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.101855,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.101855",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.14789,
      "reported_ks_context": 0.093201,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:08Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8161a699d40e209add2d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:08Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:08Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:08Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.101855,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8161-a699-d40e209add2d",
        "class": "G20",
        "model_evaluation": "s2k-cooldown-s92200",
        "wws21_mean": 0.101855,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:08Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s92200_GOOG_3855694/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s92200_GOOG_3855694/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:08Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8161a699d40e209add2d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 817,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/92200",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s92200_GOOG_3855694.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 92200)",
            "checkpoint_step": 92200,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s92200_GOOG_3855694.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-cooldown-s92200_GOOG_3855694\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 92200         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s92200_GOOG_3855694.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-cooldown-s92200_GOOG_3855694\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 92200         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s92200_GOOG_3855694.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s92200_GOOG_3855694.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-cooldown-s92200_GOOG_3855694"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s92200_GOOG_3855694/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s92200_GOOG_3855694/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s92200_GOOG_3855694/data_real.tar",
            "selector": null,
            "source_value": 0.101855,
            "report_record_id": "canonical-scores:37712c45-68fd-8161-a699-d40e209add2d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8163-8f34-e1561f72dc35",
      "source_row_id": "37712c45-68fd-8163-8f34-e1561f72dc35",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81638f34e1561f72dc35",
      "run": "H1-120M-L100-init",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.146719,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.146719",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.117506,
      "reported_ks_context": 0.079962,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:18Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81638f34e1561f72dc35",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:18Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:18Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:18Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.146719,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8163-8f34-e1561f72dc35",
        "class": "G20",
        "model_evaluation": "H1-120M-L100-init",
        "wws21_mean": 0.146719,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:18Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-L100-init/scores/scores_uncond_GOOG_H1-120M-L100-init_shard6_integrated_2655006.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-L100-init_GOOG_2655006/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:18Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-L100-init_GOOG_2655006/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81638f34e1561f72dc35",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 818,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-L100-init_GOOG_2655006/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-L100-init_GOOG_2655006.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-120M-L100-init_GOOG_2655006\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 135458         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-L100-init_GOOG_2655006.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-L100-init/scores/scores_uncond_GOOG_H1-120M-L100-init_shard6_integrated_2655006.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-L100-init/scores/scores_uncond_GOOG_H1-120M-L100-init_integrated_2655006.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14671924457613597,
        "feature_points": {
          "spread": 0.2289496329144231,
          "ask_cancellation_depth": 0.3164603773500208,
          "bid_cancellation_depth": 0.3216479509312503,
          "limit_ask_order_levels": 0.0953683807905322,
          "limit_bid_order_levels": 0.1140499789118479,
          "ask_cancellation_levels": 0.0908632060081641,
          "bid_cancellation_levels": 0.08970227625117835,
          "vol_per_min": 0.0902956914435651,
          "ofi": 0.05659427029342866,
          "ofi_up": 0.03486078547374936,
          "ofi_stay": 0.057741637320376286,
          "orderbook_imbalance": 0.11367099659798438,
          "ofi_down": 0.030869744553023167,
          "log_inter_arrival_time": 0.06212798162958446,
          "log_time_to_cancel": 0.36249344949810236,
          "ask_volume_touch": 0.05274926141578323,
          "bid_volume_touch": 0.06153720284327368,
          "ask_volume": 0.10998655450421595,
          "bid_volume": 0.12151700672438998,
          "limit_ask_order_depth": 0.3276942371600875,
          "limit_bid_order_depth": 0.3419235134838746
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-L100-init/scores/scores_uncond_GOOG_H1-120M-L100-init_integrated_2655006.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.146719,
            "report_record_id": "canonical-scores:37712c45-68fd-8163-8f34-e1561f72dc35",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8163-a68f-e7e4e0fa3306",
      "source_row_id": "37712c45-68fd-8163-a68f-e7e4e0fa3306",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8163a68fe7e4e0fa3306",
      "run": "soup_gdn_last5_gen1000",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.19253,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.19253",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.144605,
      "reported_ks_context": 0.102937,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:32:58Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8163a68fe7e4e0fa3306",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:32:58Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:32:58Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:32:58Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.19253,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8163-a68f-e7e4e0fa3306",
        "class": "G20",
        "model_evaluation": "soup_gdn_last5_gen1000",
        "wws21_mean": 0.19253,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:32:58Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen1000_GOOG_3150405/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen1000_GOOG_3150405/data_real.tar;/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen1000_GOOG_3150236/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:32:58Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3137 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8163a68fe7e4e0fa3306",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 819,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen1000_GOOG_3150405.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen1000_GOOG_3150405.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup_gdn_last5_gen1000_GOOG_3150405\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 3137         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262991+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134585+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 1000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen1000_GOOG_3150405.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup_gdn_last5_gen1000_GOOG_3150405\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 3137         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen1000_GOOG_3150405.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen1000_GOOG_3150405.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup_gdn_last5_gen1000_GOOG_3150405"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen1000_GOOG_3150405/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen1000_GOOG_3150405/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen1000_GOOG_3150405/data_real.tar",
            "selector": null,
            "source_value": 0.19253,
            "report_record_id": "canonical-scores:37712c45-68fd-8163-a68f-e7e4e0fa3306",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8164-87e6-e3ceaec0f8e5",
      "source_row_id": "37712c45-68fd-8164-87e6-e3ceaec0f8e5",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd816487e6e3ceaec0f8e5",
      "run": "R1-s4k-d256-avg3-s52-54k",
      "stock": null,
      "model_size": null,
      "run_family": "R1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.12872,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.12872",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.164252,
      "reported_ks_context": 0.094971,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:29Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd816487e6e3ceaec0f8e5",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:29Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:29Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:29Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.12872,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8164-87e6-e3ceaec0f8e5",
        "class": "G20",
        "model_evaluation": "R1-s4k-d256-avg3-s52-54k",
        "wws21_mean": 0.12872,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:29Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-avg3-s52-54k_GOOG_3745390/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-avg3-s52-54k_GOOG_3745390/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:29Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd816487e6e3ceaec0f8e5",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 820,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-avg3-s52-54k_GOOG_3745390.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292 (step 999999)",
            "checkpoint_step": 999999,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-avg3-s52-54k_GOOG_3745390.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-avg3-s52-54k_GOOG_3745390\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.163252+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-avg3-s52-54k_GOOG_3745390.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-avg3-s52-54k_GOOG_3745390\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-avg3-s52-54k_GOOG_3745390.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-avg3-s52-54k_GOOG_3745390.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-avg3-s52-54k_GOOG_3745390"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-avg3-s52-54k_GOOG_3745390/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-avg3-s52-54k_GOOG_3745390/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-avg3-s52-54k_GOOG_3745390/data_real.tar",
            "selector": null,
            "source_value": 0.12872,
            "report_record_id": "canonical-scores:37712c45-68fd-8164-87e6-e3ceaec0f8e5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8164-a50d-de27ed0655d1",
      "source_row_id": "37712c45-68fd-8164-a50d-de27ed0655d1",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8164a50dde27ed0655d1",
      "run": "pareto-360m-muon-s8606",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.241636,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.241636",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.176442,
      "reported_ks_context": 0.137775,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:02Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8164a50dde27ed0655d1",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:02Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:02Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:02Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.241636,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8164-a50d-de27ed0655d1",
        "class": "G20",
        "model_evaluation": "pareto-360m-muon-s8606",
        "wws21_mean": 0.241636,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:02Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s8606_GOOG_3253410/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s8606_GOOG_3253410/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:02Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8164a50dde27ed0655d1",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 821,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s8606_GOOG_3253410.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592 (step 8606)",
            "checkpoint_step": 8606,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s8606_GOOG_3253410.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s8606_GOOG_3253410\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 8606         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263086+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158865+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s8606_GOOG_3253410.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s8606_GOOG_3253410\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 8606         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s8606_GOOG_3253410.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s8606_GOOG_3253410.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s8606_GOOG_3253410"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s8606_GOOG_3253410/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s8606_GOOG_3253410/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s8606_GOOG_3253410/data_real.tar",
            "selector": null,
            "source_value": 0.241636,
            "report_record_id": "canonical-scores:37712c45-68fd-8164-a50d-de27ed0655d1",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8166-b135-c51d41369370",
      "source_row_id": "37712c45-68fd-8166-b135-c51d41369370",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8166b135c51d41369370",
      "run": "pareto-120m-s58990",
      "stock": null,
      "model_size": "120M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.194815,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.194815",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.145776,
      "reported_ks_context": 0.103097,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:02Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8166b135c51d41369370",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:02Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:02Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:02Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.194815,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8166-b135-c51d41369370",
        "class": "G20",
        "model_evaluation": "pareto-120m-s58990",
        "wws21_mean": 0.194815,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:02Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s58990_GOOG_3253396/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s58990_GOOG_3253396/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:02Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8166b135c51d41369370",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 822,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292/58990",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s58990_GOOG_3253396.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292 (step 58990)",
            "checkpoint_step": 58990,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s58990_GOOG_3253396.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s58990_GOOG_3253396\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 58990         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.165112+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s58990_GOOG_3253396.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s58990_GOOG_3253396\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 58990         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s58990_GOOG_3253396.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s58990_GOOG_3253396.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s58990_GOOG_3253396"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s58990_GOOG_3253396/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s58990_GOOG_3253396/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s58990_GOOG_3253396/data_real.tar",
            "selector": null,
            "source_value": 0.194815,
            "report_record_id": "canonical-scores:37712c45-68fd-8166-b135-c51d41369370",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8166-b923-c21f1b80b5ec",
      "source_row_id": "37712c45-68fd-8166-b923-c21f1b80b5ec",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8166b923c21f1b80b5ec",
      "run": "adamw-s39275",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.149571,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.149571",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.117263,
      "reported_ks_context": 0.084166,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:47Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8166b923c21f1b80b5ec",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:47Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:47Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:47Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.149571,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8166-b923-c21f1b80b5ec",
        "class": "G20",
        "model_evaluation": "adamw-s39275",
        "wws21_mean": 0.149571,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:47Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s39275_GOOG_2868484/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s39275_GOOG_2868484/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:47Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8166b923c21f1b80b5ec",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 823,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s39275_GOOG_2868484.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665 (step 39275)",
            "checkpoint_step": 39275,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s39275_GOOG_2868484.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s39275_GOOG_2868484\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 39275         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262928+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.121688+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s39275_GOOG_2868484.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s39275_GOOG_2868484\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 39275         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s39275_GOOG_2868484.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s39275_GOOG_2868484.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-s39275_GOOG_2868484"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s39275_GOOG_2868484/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s39275_GOOG_2868484/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s39275_GOOG_2868484/data_real.tar",
            "selector": null,
            "source_value": 0.149571,
            "report_record_id": "canonical-scores:37712c45-68fd-8166-b923-c21f1b80b5ec",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8168-83ce-e08271c4d36f",
      "source_row_id": "37712c45-68fd-8168-83ce-e08271c4d36f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd816883cee08271c4d36f",
      "run": "j2505830_pfr",
      "stock": null,
      "model_size": null,
      "run_family": "j2505830",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.214499,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.214499",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.158616,
      "reported_ks_context": 0.113271,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-28T06:57:57Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2163",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-28T06:57:57Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-28T06:57:57Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-28T06:57:57Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.214499,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8168-83ce-e08271c4d36f",
        "class": "G20",
        "model_evaluation": "j2505830_pfr",
        "wws21_mean": 0.214499,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-28T06:57:57Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-28T06:57:57Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd816883cee08271c4d36f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 824,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_pfr_GOOG_3152214.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830 (step 38349)",
            "checkpoint_step": 38349,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_pfr_GOOG_3152214.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 38349         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.166304+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_pfr_GOOG_3152214.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 38349         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_pfr_GOOG_3152214.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_pfr_GOOG_3152214.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214/data_real.tar",
            "selector": null,
            "source_value": 0.214499,
            "report_record_id": "canonical-scores:37712c45-68fd-8168-83ce-e08271c4d36f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8169-884d-faf9e23e7d20",
      "source_row_id": "37712c45-68fd-8169-884d-faf9e23e7d20",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8169884dfaf9e23e7d20",
      "run": "ticks-55m-rand",
      "stock": null,
      "model_size": "55M",
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.223094,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.223094",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.171242,
      "reported_ks_context": 0.115091,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:06Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8169884dfaf9e23e7d20",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:06Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:06Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:06Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.223094,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8169-884d-faf9e23e7d20",
        "class": "G20",
        "model_evaluation": "ticks-55m-rand",
        "wws21_mean": 0.223094,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:06Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-55m-rand/scores/scores_uncond_GOOG_ticks-55m-rand_integrated_3278221.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:06Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8169884dfaf9e23e7d20",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 825,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617/log_rank0.txt",
            "evidence_line": 634,
            "evidence_text": "[Checkpoint] Loading step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-55m-rand_GOOG_3278221.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-55m-rand_GOOG_3278221.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-55m-rand_GOOG_3278221.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-55m-rand_integrated_3278221.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-55m-rand_GOOG_2887617.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-55m-rand_GOOG_2887617.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-55m-rand_GOOG_3278221.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-55m-rand_GOOG_3278221.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-55m-rand_GOOG_2887617"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-55m-rand_GOOG_3278221.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-55m-rand_integrated_3278221.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-55m-rand/scores/scores_uncond_GOOG_ticks-55m-rand_integrated_3278221.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-55m-rand/scores/scores_uncond_GOOG_ticks-55m-rand_integrated_3278221.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.22309377333810326,
        "feature_points": {
          "ofi_up": 0.08203170568861654,
          "ofi_stay": 0.05238323024791695,
          "spread": 0.3042915771654641,
          "ask_cancellation_depth": 0.33860412757828656,
          "bid_cancellation_ticks": 0.3433577233457224,
          "bid_cancellation_depth": 0.34410491504480195,
          "ofi_down": 0.0709777175037117,
          "ask_cancellation_ticks": 0.33800261020929107,
          "limit_bid_order_depth": 0.357629773189992,
          "ask_volume": 0.21780447215181653,
          "vol_per_min": 0.09584228563871938,
          "limit_ask_order_ticks": 0.3487214667951836,
          "limit_ask_order_depth": 0.3486699645177928,
          "ask_volume_touch": 0.1117383134103259,
          "log_time_to_cancel": 0.3997914588069087,
          "limit_bid_order_ticks": 0.3575762781899478,
          "bid_volume": 0.2626913159755645,
          "bid_volume_touch": 0.08925932777767114,
          "orderbook_imbalance": 0.1135731873204866,
          "ofi": 0.0511938899408996,
          "log_inter_arrival_time": 0.056723899601049066
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-55m-rand/scores/scores_uncond_GOOG_ticks-55m-rand_integrated_3278221.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.223094,
            "report_record_id": "canonical-scores:37712c45-68fd-8169-884d-faf9e23e7d20",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-816b-880f-e44d7828227e",
      "source_row_id": "37712c45-68fd-816b-880f-e44d7828227e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd816b880fe44d7828227e",
      "run": "ctx4k-cond1000",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.160984,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.160984",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.126431,
      "reported_ks_context": 0.088901,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:59Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd816b880fe44d7828227e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:59Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:59Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:59Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.160984,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-816b-880f-e44d7828227e",
        "class": "G20",
        "model_evaluation": "ctx4k-cond1000",
        "wws21_mean": 0.160984,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:59Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond1000/scores/scores_uncond_GOOG_ctx4k-cond1000_integrated_2828760.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond1000_GOOG_2828760/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:59Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond1000_GOOG_2828760/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd816b880fe44d7828227e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 826,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond1000_GOOG_2828760/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond1000_GOOG_2828760.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond1000_GOOG_2828760\"         --n_cond_msgs 1000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond1000_GOOG_2828760.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond1000/scores/scores_uncond_GOOG_ctx4k-cond1000_integrated_2828760.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond1000/scores/scores_uncond_GOOG_ctx4k-cond1000_integrated_2828760.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.16098402509107965,
        "feature_points": {
          "ask_cancellation_depth": 0.27294669973375074,
          "limit_bid_order_levels": 0.0187345805432504,
          "ofi_down": 0.07747312419097593,
          "vol_per_min": 0.16515522052586246,
          "limit_ask_order_depth": 0.33875549124250376,
          "ask_volume": 0.22457146653944657,
          "bid_cancellation_levels": 0.022841181273543613,
          "ofi_up": 0.03856007330495829,
          "ask_cancellation_levels": 0.04721622999294265,
          "ask_volume_touch": 0.12304481425972535,
          "limit_ask_order_levels": 0.05022109545589527,
          "orderbook_imbalance": 0.09180431676620954,
          "bid_volume": 0.3298874402778669,
          "spread": 0.27406579645294327,
          "limit_bid_order_depth": 0.2948112239397239,
          "ofi_stay": 0.07881901589520218,
          "bid_cancellation_depth": 0.35008394083880146,
          "log_inter_arrival_time": 0.061078631994923534,
          "bid_volume_touch": 0.05738835001285443,
          "log_time_to_cancel": 0.38645446288074903,
          "ofi": 0.0767513707905431
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond1000/scores/scores_uncond_GOOG_ctx4k-cond1000_integrated_2828760.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.160984,
            "report_record_id": "canonical-scores:37712c45-68fd-816b-880f-e44d7828227e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-816b-9745-c2fe35437008",
      "source_row_id": "37712c45-68fd-816b-9745-c2fe35437008",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd816b9745c2fe35437008",
      "run": "soup_gdn_last5_gen2000",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.182183,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.182183",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.160539,
      "reported_ks_context": 0.119536,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:43Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd816b9745c2fe35437008",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:43Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:43Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:43Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.182183,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-816b-9745-c2fe35437008",
        "class": "G20",
        "model_evaluation": "soup_gdn_last5_gen2000",
        "wws21_mean": 0.182183,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:43Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_soup_gdn_last5_gen2000/scores/scores_uncond_GOOG_soup_gdn_last5_gen2000_integrated_3150406.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen2000_GOOG_3150406/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:43Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 3137 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen2000_GOOG_3150406/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd816b9745c2fe35437008",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 827,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5/0",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup_gdn_last5_gen2000_GOOG_3150406/log_rank0.txt",
            "evidence_line": 702,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5/0/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5/0",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5/0",
            "availability": "not_found",
            "path_scope": "checkpoint_step",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 2000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen2000_GOOG_3150406.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup_gdn_last5_gen2000_GOOG_3150406\"         --n_cond_msgs 500         --n_gen_msgs 2000         --n_sequences 3137         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup_gdn_last5_gen2000_GOOG_3150406.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_soup_gdn_last5_gen2000/scores/scores_uncond_GOOG_soup_gdn_last5_gen2000_integrated_3150406.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_soup_gdn_last5_gen2000/scores/scores_uncond_GOOG_soup_gdn_last5_gen2000_integrated_3150406.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.18218276526536906,
        "feature_points": {
          "limit_bid_order_levels": 0.2407105796442102,
          "vol_per_min": 0.0569476547415142,
          "ofi_up": 0.03502443763381481,
          "ask_cancellation_levels": 0.22111401404988915,
          "ofi_down": 0.043173140293165164,
          "bid_cancellation_levels": 0.20193803011347822,
          "limit_ask_order_levels": 0.26345291172936114,
          "bid_cancellation_depth": 0.38209759257051,
          "limit_bid_order_depth": 0.3651404595222592,
          "limit_ask_order_depth": 0.12992414469084274,
          "ask_cancellation_depth": 0.3565003518700006,
          "log_time_to_cancel": 0.3417459053150445,
          "ofi_stay": 0.09546572978841966,
          "orderbook_imbalance": 0.08723337283548624,
          "ofi": 0.09428647214329239,
          "log_inter_arrival_time": 0.12102370051728777,
          "bid_volume": 0.14855771578825258,
          "spread": 0.3920880961093915,
          "bid_volume_touch": 0.0827347568811289,
          "ask_volume": 0.105488428074465,
          "ask_volume_touch": 0.06119057626093629
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_soup_gdn_last5_gen2000/scores/scores_uncond_GOOG_soup_gdn_last5_gen2000_integrated_3150406.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.182183,
            "report_record_id": "canonical-scores:37712c45-68fd-816b-9745-c2fe35437008",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-816b-b436-f2f0c7f916ff",
      "source_row_id": "37712c45-68fd-816b-b436-f2f0c7f916ff",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd816bb436f2f0c7f916ff",
      "run": "H2-55M-ctx8k-best",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.173543,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.173543",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.132925,
      "reported_ks_context": 0.095983,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:10Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd816bb436f2f0c7f916ff",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:10Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:10Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:10Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.173543,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-816b-b436-f2f0c7f916ff",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx8k-best",
        "wws21_mean": 0.173543,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:10Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-best/scores/scores_uncond_GOOG_H2-55M-ctx8k-best_integrated_2572656.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-best_GOOG_2572656/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:10Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-best_GOOG_2572656/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd816bb436f2f0c7f916ff",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 828,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "58091",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-best_GOOG_2572656/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "41850",
                "45450",
                "46650",
                "47850",
                "49050",
                "50250",
                "51450",
                "52650",
                "53849",
                "55048",
                "56248",
                "57434",
                "58091",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx8k-best_GOOG_2572656.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-best_GOOG_2572656\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 58091         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx8k-best_GOOG_2572656.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-best/scores/scores_uncond_GOOG_H2-55M-ctx8k-best_integrated_2572656.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-best/scores/scores_uncond_GOOG_H2-55M-ctx8k-best_integrated_2572656.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.17354273735997325,
        "feature_points": {
          "spread": 0.38066209592697464,
          "ask_cancellation_depth": 0.30366976759541703,
          "bid_cancellation_depth": 0.32345716749692033,
          "limit_ask_order_levels": 0.04331410365480864,
          "limit_bid_order_levels": 0.07460208897870128,
          "ask_cancellation_levels": 0.12048029799161782,
          "bid_cancellation_levels": 0.12014224702687809,
          "vol_per_min": 0.1424067537921066,
          "ofi": 0.10597656292803774,
          "ofi_up": 0.13960797679644757,
          "ofi_stay": 0.10527862961295957,
          "orderbook_imbalance": 0.10405945987061202,
          "ofi_down": 0.10405578438435784,
          "log_inter_arrival_time": 0.013627229318948773,
          "log_time_to_cancel": 0.39203642534043176,
          "ask_volume_touch": 0.12377636448545726,
          "bid_volume_touch": 0.09916271832073814,
          "ask_volume": 0.12386485954652089,
          "bid_volume": 0.18484104490132255,
          "limit_ask_order_depth": 0.30598082519982633,
          "limit_bid_order_depth": 0.3333950813903538
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-best/scores/scores_uncond_GOOG_H2-55M-ctx8k-best_integrated_2572656.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.173543,
            "report_record_id": "canonical-scores:37712c45-68fd-816b-b436-f2f0c7f916ff",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-816c-abfc-da4d280948a2",
      "source_row_id": "37712c45-68fd-816c-abfc-da4d280948a2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd816cabfcda4d280948a2",
      "run": "mamba3-paperbase-pw8u-s46050-c250g250-absmax8k",
      "stock": null,
      "model_size": null,
      "run_family": "mamba3",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 1.134503,
      "source_value_field": "Wasserstein",
      "source_value_text": "1.134503",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.781039,
      "reported_ks_context": 0.65616,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:14Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd816cabfcda4d280948a2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:14Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:14Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:14Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 1.134503,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-816c-abfc-da4d280948a2",
        "class": "G20",
        "model_evaluation": "mamba3-paperbase-pw8u-s46050-c250g250-absmax8k",
        "wws21_mean": 1.134503,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:14Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:14Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd816cabfcda4d280948a2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 829,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629/46050",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629 (step 46050)",
            "checkpoint_step": 46050,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 3136         --batch_size 64         --checkpoint_step 46050         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136_absmax8000_as_s500.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.166575+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 3136         --batch_size 64         --checkpoint_step 46050         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136_absmax8000_as_s500.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
            "selector": null,
            "source_value": 1.134503,
            "report_record_id": "canonical-scores:37712c45-68fd-816c-abfc-da4d280948a2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-816c-b9bb-e91d09bdb300",
      "source_row_id": "37712c45-68fd-816c-b9bb-e91d09bdb300",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd816cb9bbe91d09bdb300",
      "run": "ctx1k-cond500",
      "stock": null,
      "model_size": null,
      "run_family": "ctx1k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.177865,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.177865",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.132876,
      "reported_ks_context": 0.099835,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:06Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd816cb9bbe91d09bdb300",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:06Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:06Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:06Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.177865,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-816c-b9bb-e91d09bdb300",
        "class": "G20",
        "model_evaluation": "ctx1k-cond500",
        "wws21_mean": 0.177865,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:06Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx1k-cond500/scores/scores_uncond_GOOG_ctx1k-cond500_integrated_2826361.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx1k-cond500_GOOG_2826361/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:06Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx1k-cond500_GOOG_2826361/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd816cb9bbe91d09bdb300",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 830,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "93205",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx1k-cond500_GOOG_2826361/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=93205 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230/93205",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "32395",
                "55635",
                "62605",
                "74222",
                "76545",
                "78868",
                "81191",
                "83493",
                "85811",
                "88136",
                "90459",
                "9150",
                "92785",
                "93205",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230/93205",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx1k-cond500_GOOG_2826361.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-1k/checkpoints/j2781230_bsj3f4zh_2781230\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx1k-cond500_GOOG_2826361\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 93205         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx1k-cond500_GOOG_2826361.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx1k-cond500/scores/scores_uncond_GOOG_ctx1k-cond500_integrated_2826361.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx1k-cond500/scores/scores_uncond_GOOG_ctx1k-cond500_integrated_2826361.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.17786534004910304,
        "feature_points": {
          "limit_ask_order_depth": 0.3060836662798709,
          "ofi_up": 0.11634665699377039,
          "limit_ask_order_levels": 0.07085374270884201,
          "ask_volume_touch": 0.10608494144594409,
          "ofi_down": 0.11291324267155042,
          "bid_cancellation_depth": 0.3268828541709721,
          "log_time_to_cancel": 0.3305962965685544,
          "vol_per_min": 0.1326341913141405,
          "ofi_stay": 0.08586547339607879,
          "spread": 0.26254338409713596,
          "limit_bid_order_levels": 0.09870855650178184,
          "bid_cancellation_levels": 0.10072864511974887,
          "ask_cancellation_levels": 0.1215055727616778,
          "bid_volume_touch": 0.1381825815846135,
          "ask_cancellation_depth": 0.31632115987360776,
          "bid_volume": 0.29009855088813996,
          "ofi": 0.0870394327224418,
          "orderbook_imbalance": 0.1056004342710725,
          "limit_bid_order_depth": 0.3401616543973376,
          "ask_volume": 0.2449824692173639,
          "log_inter_arrival_time": 0.04103863404651894
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx1k-cond500/scores/scores_uncond_GOOG_ctx1k-cond500_integrated_2826361.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.177865,
            "report_record_id": "canonical-scores:37712c45-68fd-816c-b9bb-e91d09bdb300",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-816c-b9fe-d29fa05f310a",
      "source_row_id": "37712c45-68fd-816c-b9fe-d29fa05f310a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd816cb9fed29fa05f310a",
      "run": "phaseB-m3-14m-s1000-d256-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.122009,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.122009",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.169766,
      "reported_ks_context": 0.103713,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:14Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd816cb9fed29fa05f310a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:14Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:14Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:14Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.122009,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-816c-b9fe-d29fa05f310a",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s1000-d256-b-avg-c500g500",
        "wws21_mean": 0.122009,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:14Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:14Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd816cb9fed29fa05f310a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 831,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932/log_rank0.txt",
            "evidence_line": 651,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704412_bmp7l150_3704412",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704412_bmp7l150_3704412\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704412_bmp7l150_3704412 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932/log_rank0.txt",
                "evidence_line": 653,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412",
            "availability_checked_at": "2026-09-12T06:13:15.375258+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704412_bmp7l150_3704412/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704412_bmp7l150_3704412\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d256-b-avg-c500g500_GOOG_3713932/data_real.tar",
            "selector": null,
            "source_value": 0.122009,
            "report_record_id": "canonical-scores:37712c45-68fd-816c-b9fe-d29fa05f310a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-816f-b929-f1774a561c8a",
      "source_row_id": "37712c45-68fd-816f-b929-f1774a561c8a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd816fb929f1774a561c8a",
      "run": "muon-s70903",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.170055,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.170055",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.129179,
      "reported_ks_context": 0.084486,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd816fb929f1774a561c8a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.170055,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-816f-b929-f1774a561c8a",
        "class": "G20",
        "model_evaluation": "muon-s70903",
        "wws21_mean": 0.170055,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s70903_GOOG_2868873/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s70903_GOOG_2868873/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd816fb929f1774a561c8a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 832,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s70903_GOOG_2868873.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925 (step 70903)",
            "checkpoint_step": 70903,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s70903_GOOG_2868873.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s70903_GOOG_2868873\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 70903         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263108+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.161994+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s70903_GOOG_2868873.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782925_4icb4jua_2782925\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s70903_GOOG_2868873\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 70903         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s70903_GOOG_2868873.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s70903_GOOG_2868873.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-s70903_GOOG_2868873"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s70903_GOOG_2868873/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s70903_GOOG_2868873/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s70903_GOOG_2868873/data_real.tar",
            "selector": null,
            "source_value": 0.170055,
            "report_record_id": "canonical-scores:37712c45-68fd-816f-b929-f1774a561c8a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8170-9d04-e90de519925f",
      "source_row_id": "37712c45-68fd-8170-9d04-e90de519925f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81709d04e90de519925f",
      "run": "pareto-360m-muon-s4407",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.269867,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.269867",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.201985,
      "reported_ks_context": 0.131477,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:31Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81709d04e90de519925f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:31Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:31Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:31Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.269867,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8170-9d04-e90de519925f",
        "class": "G20",
        "model_evaluation": "pareto-360m-muon-s4407",
        "wws21_mean": 0.269867,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:31Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s4407_GOOG_3253409/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s4407_GOOG_3253409/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:31Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81709d04e90de519925f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 833,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s4407_GOOG_3253409.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592 (step 4407)",
            "checkpoint_step": 4407,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s4407_GOOG_3253409.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s4407_GOOG_3253409\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 4407         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263086+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158865+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s4407_GOOG_3253409.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s4407_GOOG_3253409\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 4407         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s4407_GOOG_3253409.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s4407_GOOG_3253409.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s4407_GOOG_3253409"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s4407_GOOG_3253409/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s4407_GOOG_3253409/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s4407_GOOG_3253409/data_real.tar",
            "selector": null,
            "source_value": 0.269867,
            "report_record_id": "canonical-scores:37712c45-68fd-8170-9d04-e90de519925f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8174-921e-c1a6aee8b521",
      "source_row_id": "37712c45-68fd-8174-921e-c1a6aee8b521",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8174921ec1a6aee8b521",
      "run": "j2504167-step100378-L100-c500",
      "stock": null,
      "model_size": null,
      "run_family": "j2504167",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.161248,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.161248",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.125725,
      "reported_ks_context": 0.088191,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T21:58:46Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2180",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T21:58:46Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T21:58:46Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T21:58:46Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.161248,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8174-921e-c1a6aee8b521",
        "class": "G20",
        "model_evaluation": "j2504167-step100378-L100-c500",
        "wws21_mean": 0.161248,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T21:58:46Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167-step100378-L100-c500/scores/scores_uncond_GOOG_j2504167-step100378-L100-c500_integrated_2757353.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167-step100378-L100-c500_GOOG_2757353/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T21:58:46Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167-step100378-L100-c500_GOOG_2757353/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8174921ec1a6aee8b521",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 834,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167-step100378-L100-c500_GOOG_2757353/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=100378 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504167-step100378-L100-c500_GOOG_2757353.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504167-step100378-L100-c500_GOOG_2757353\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 100378         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504167-step100378-L100-c500_GOOG_2757353.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167-step100378-L100-c500/scores/scores_uncond_GOOG_j2504167-step100378-L100-c500_integrated_2757353.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167-step100378-L100-c500/scores/scores_uncond_GOOG_j2504167-step100378-L100-c500_integrated_2757353.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.16124822272496397,
        "feature_points": {
          "limit_bid_order_levels": 0.041140950055010554,
          "orderbook_imbalance": 0.10012002660079009,
          "ofi_up": 0.09963108455036503,
          "ofi_stay": 0.08437048740998893,
          "ask_cancellation_levels": 0.059595830770028016,
          "spread": 0.23670595986794155,
          "log_time_to_cancel": 0.38762420471924575,
          "limit_bid_order_depth": 0.3623299147845962,
          "vol_per_min": 0.09345311014077745,
          "bid_cancellation_levels": 0.049883849033818733,
          "ask_cancellation_depth": 0.329896892391412,
          "ofi": 0.08401990722952535,
          "limit_ask_order_levels": 0.025526563108358973,
          "limit_ask_order_depth": 0.34742875522695055,
          "ask_volume_touch": 0.10861253831733911,
          "ofi_down": 0.04276078466317555,
          "bid_volume": 0.2510560567646989,
          "bid_cancellation_depth": 0.3569087325796218,
          "bid_volume_touch": 0.08176205459515844,
          "log_inter_arrival_time": 0.05194878405491306,
          "ask_volume": 0.1914361903605274
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167-step100378-L100-c500/scores/scores_uncond_GOOG_j2504167-step100378-L100-c500_integrated_2757353.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.161248,
            "report_record_id": "canonical-scores:37712c45-68fd-8174-921e-c1a6aee8b521",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8175-96d4-fce8023a6077",
      "source_row_id": "37712c45-68fd-8175-96d4-fce8023a6077",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd817596d4fce8023a6077",
      "run": "H1_120M_step10457",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.209162,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.209162",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.163004,
      "reported_ks_context": 0.109012,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:16Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd817596d4fce8023a6077",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:16Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:16Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:16Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.209162,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8175-96d4-fce8023a6077",
        "class": "G20",
        "model_evaluation": "H1_120M_step10457",
        "wws21_mean": 0.209162,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:16Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_120M_step10457/scores/scores_uncond_GOOG_H1_120M_step10457_shard20_integrated_2566029.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1_120M_step10457_GOOG_2566029/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:16Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1_120M_step10457_GOOG_2566029/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd817596d4fce8023a6077",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 835,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "10457",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260/10457",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1_120M_step10457_GOOG_2566029/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260/10457/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260/10457",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260/10457",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "10457",
                "12586",
                "14708",
                "16827",
                "18953",
                "21080",
                "23207",
                "4072",
                "6198",
                "8330",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260/10457",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1_120M_step10457_GOOG_2566029.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1_120M_step10457_GOOG_2566029\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 10457         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1_120M_step10457_GOOG_2566029.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_120M_step10457/scores/scores_uncond_GOOG_H1_120M_step10457_shard20_integrated_2566029.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_120M_step10457/scores/scores_uncond_GOOG_H1_120M_step10457_integrated_2566029.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.2091622019553317,
        "feature_points": {
          "spread": 0.3716786791121821,
          "ask_cancellation_depth": 0.3947994852132295,
          "bid_cancellation_depth": 0.38095980749514957,
          "limit_ask_order_levels": 0.11392430901293085,
          "limit_bid_order_levels": 0.11242553202103524,
          "ask_cancellation_levels": 0.1822246882318129,
          "bid_cancellation_levels": 0.19537277858476923,
          "vol_per_min": 0.15232850422830002,
          "ofi": 0.07911891771163966,
          "ofi_up": 0.13959843211657766,
          "ofi_stay": 0.07833860753995071,
          "orderbook_imbalance": 0.08912752504496745,
          "ofi_down": 0.12429619126141482,
          "log_inter_arrival_time": 0.0984417239703699,
          "log_time_to_cancel": 0.5909968880141749,
          "ask_volume_touch": 0.12843089359828452,
          "bid_volume_touch": 0.14254767412982294,
          "ask_volume": 0.11089351948265232,
          "bid_volume": 0.12116723857922974,
          "limit_ask_order_depth": 0.40260833861911316,
          "limit_bid_order_depth": 0.38312650709435836
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1_120M_step10457/scores/scores_uncond_GOOG_H1_120M_step10457_integrated_2566029.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.209162,
            "report_record_id": "canonical-scores:37712c45-68fd-8175-96d4-fce8023a6077",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8177-8e4e-e11b62641976",
      "source_row_id": "37712c45-68fd-8177-8e4e-e11b62641976",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81778e4ee11b62641976",
      "run": "H2-55M-ctx4k-cond5000",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.144629,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.144629",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.113743,
      "reported_ks_context": 0.076617,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:30Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81778e4ee11b62641976",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:30Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:30Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:30Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.144629,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8177-8e4e-e11b62641976",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond5000",
        "wws21_mean": 0.144629,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:30Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond5000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond5000_shard13_integrated_2632448.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond5000_GOOG_2632448/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:30Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond5000_GOOG_2632448/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81778e4ee11b62641976",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 836,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond5000_GOOG_2632448/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 5000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond5000_GOOG_2632448.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond5000_GOOG_2632448\"         --n_cond_msgs 5000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 6         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond5000_GOOG_2632448.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond5000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond5000_shard13_integrated_2632448.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond5000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond5000_integrated_2632448.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14462871042761843,
        "feature_points": {
          "spread": 0.17453661369688095,
          "ask_cancellation_depth": 0.3410820683592653,
          "bid_cancellation_depth": 0.34200003218269864,
          "limit_ask_order_levels": 0.01964795344895571,
          "limit_bid_order_levels": 0.025141877495134135,
          "ask_cancellation_levels": 0.029396532941637882,
          "bid_cancellation_levels": 0.04178460746078423,
          "vol_per_min": 0.12196729043519845,
          "ofi": 0.038115673743614395,
          "ofi_up": 0.09281644621282174,
          "ofi_stay": 0.03856630767020179,
          "orderbook_imbalance": 0.10470719379175995,
          "ofi_down": 0.0757641658992875,
          "log_inter_arrival_time": 0.03212958452781767,
          "log_time_to_cancel": 0.37323520482057154,
          "ask_volume_touch": 0.09549376910047526,
          "bid_volume_touch": 0.14777899288485893,
          "ask_volume": 0.12508092682345254,
          "bid_volume": 0.14347893102184092,
          "limit_ask_order_depth": 0.3426463515604233,
          "limit_bid_order_depth": 0.331832394902306
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond5000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond5000_integrated_2632448.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.144629,
            "report_record_id": "canonical-scores:37712c45-68fd-8177-8e4e-e11b62641976",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8177-997b-ebe231cccfdb",
      "source_row_id": "37712c45-68fd-8177-997b-ebe231cccfdb",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8177997bebe231cccfdb",
      "run": "46m-s2k-d256-4n-lr01-step22650",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.120541,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.120541",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.158652,
      "reported_ks_context": 0.099739,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:19Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8177997bebe231cccfdb",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:19Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:19Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:19Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.120541,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8177-997b-ebe231cccfdb",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-4n-lr01-step22650",
        "wws21_mean": 0.120541,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:19Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-4n-lr01-step22650_GOOG_3825821/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-4n-lr01-step22650_GOOG_3825821/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:19Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8177997bebe231cccfdb",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 837,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596/22650",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-4n-lr01-step22650_GOOG_3825821.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596 (step 22650)",
            "checkpoint_step": 22650,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-4n-lr01-step22650_GOOG_3825821.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-4n-lr01-step22650_GOOG_3825821\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 22650         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.121699+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-4n-lr01-step22650_GOOG_3825821.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-4n-lr01-step22650_GOOG_3825821\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 22650         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-4n-lr01-step22650_GOOG_3825821.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-4n-lr01-step22650_GOOG_3825821.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-4n-lr01-step22650_GOOG_3825821"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-4n-lr01-step22650_GOOG_3825821/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-4n-lr01-step22650_GOOG_3825821/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-4n-lr01-step22650_GOOG_3825821/data_real.tar",
            "selector": null,
            "source_value": 0.120541,
            "report_record_id": "canonical-scores:37712c45-68fd-8177-997b-ebe231cccfdb",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8177-a187-d2df7b9d333c",
      "source_row_id": "37712c45-68fd-8177-a187-d2df7b9d333c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8177a187d2df7b9d333c",
      "run": "adamw-32n-s30978",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.140417,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.140417",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.108615,
      "reported_ks_context": 0.073929,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8177a187d2df7b9d333c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.140417,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8177-a187-d2df7b9d333c",
        "class": "G20",
        "model_evaluation": "adamw-32n-s30978",
        "wws21_mean": 0.140417,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30978_GOOG_2871391/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30978_GOOG_2871391/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8177a187d2df7b9d333c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 838,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30978_GOOG_2871391.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 30978)",
            "checkpoint_step": 30978,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30978_GOOG_2871391.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s30978_GOOG_2871391\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 30978         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30978_GOOG_2871391.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s30978_GOOG_2871391\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 30978         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30978_GOOG_2871391.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s30978_GOOG_2871391.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s30978_GOOG_2871391"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30978_GOOG_2871391/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30978_GOOG_2871391/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s30978_GOOG_2871391/data_real.tar",
            "selector": null,
            "source_value": 0.140417,
            "report_record_id": "canonical-scores:37712c45-68fd-8177-a187-d2df7b9d333c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8179-8c6e-eae782fedac2",
      "source_row_id": "37712c45-68fd-8179-8c6e-eae782fedac2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81798c6eeae782fedac2",
      "run": "46m-s4k-d512-step11435",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.172855,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.172855",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.266785,
      "reported_ks_context": 0.148753,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81798c6eeae782fedac2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.172855,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8179-8c6e-eae782fedac2",
        "class": "G20",
        "model_evaluation": "46m-s4k-d512-step11435",
        "wws21_mean": 0.172855,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-step11435_GOOG_3775013/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-step11435_GOOG_3775013/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81798c6eeae782fedac2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 839,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763900_9pcq1sv9_3763900/11435",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-step11435_GOOG_3775013.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763900_9pcq1sv9_3763900 (step 11435)",
            "checkpoint_step": 11435,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763900_9pcq1sv9_3763900",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-step11435_GOOG_3775013.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763900_9pcq1sv9_3763900\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s4k-d512-step11435_GOOG_3775013\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 11435         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.167987+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763900_9pcq1sv9_3763900",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763900_9pcq1sv9_3763900"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-step11435_GOOG_3775013.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763900_9pcq1sv9_3763900\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s4k-d512-step11435_GOOG_3775013\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 11435         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-step11435_GOOG_3775013.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s4k-d512-step11435_GOOG_3775013.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s4k-d512-step11435_GOOG_3775013"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-step11435_GOOG_3775013/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-step11435_GOOG_3775013/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s4k-d512-step11435_GOOG_3775013/data_real.tar",
            "selector": null,
            "source_value": 0.172855,
            "report_record_id": "canonical-scores:37712c45-68fd-8179-8c6e-eae782fedac2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-817b-b53c-fef40f89dd6f",
      "source_row_id": "37712c45-68fd-817b-b53c-fef40f89dd6f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd817bb53cfef40f89dd6f",
      "run": "muon-s40980",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.197886,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.197886",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.15312,
      "reported_ks_context": 0.095098,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:45Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd817bb53cfef40f89dd6f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:45Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:45Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:45Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.197886,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-817b-b53c-fef40f89dd6f",
        "class": "G20",
        "model_evaluation": "muon-s40980",
        "wws21_mean": 0.197886,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:45Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s40980_GOOG_2868870/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s40980_GOOG_2868870/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:45Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd817bb53cfef40f89dd6f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 840,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s40980_GOOG_2868870.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524 (step 40980)",
            "checkpoint_step": 40980,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s40980_GOOG_2868870.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s40980_GOOG_2868870\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 40980         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263138+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.169324+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s40980_GOOG_2868870.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s40980_GOOG_2868870\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 40980         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s40980_GOOG_2868870.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s40980_GOOG_2868870.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-s40980_GOOG_2868870"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s40980_GOOG_2868870/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s40980_GOOG_2868870/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s40980_GOOG_2868870/data_real.tar",
            "selector": null,
            "source_value": 0.197886,
            "report_record_id": "canonical-scores:37712c45-68fd-817b-b53c-fef40f89dd6f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-817b-b56d-ca3b4b014ce1",
      "source_row_id": "37712c45-68fd-817b-b56d-ca3b4b014ce1",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd817bb56dca3b4b014ce1",
      "run": "G10-L10-ep40",
      "stock": null,
      "model_size": null,
      "run_family": "G10",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.430853,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.430853",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.299286,
      "reported_ks_context": 0.267001,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:16Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd817bb56dca3b4b014ce1",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:16Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:16Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:16Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.430853,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-817b-b56d-ca3b4b014ce1",
        "class": "G20",
        "model_evaluation": "G10-L10-ep40",
        "wws21_mean": 0.430853,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:16Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_G10-L10-ep40/scores/scores_uncond_GOOG_G10-L10-ep40_integrated_2789708.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/G10-L10-ep40_GOOG_2789708/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:16Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/G10-L10-ep40_GOOG_2789708/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd817bb56dca3b4b014ce1",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 841,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "95120",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2782577_m9jc2jt5_2782577",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2782577_m9jc2jt5_2782577",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/G10-L10-ep40_GOOG_2789708/log_rank0.txt",
            "evidence_line": 8,
            "evidence_text": "[Checkpoint] Loading step=95120 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2782577_m9jc2jt5_2782577 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2782577_m9jc2jt5_2782577",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2782577_m9jc2jt5_2782577/95120",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "11890",
                "23780",
                "35670",
                "47560",
                "59450",
                "71340",
                "73718",
                "76096",
                "78474",
                "80852",
                "83230",
                "85608",
                "87986",
                "90364",
                "92742",
                "95120",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2782577_m9jc2jt5_2782577/95120",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-L10-ep40_GOOG_2789708.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2782577_m9jc2jt5_2782577\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/G10-L10-ep40_GOOG_2789708\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 95120         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-L10-ep40_GOOG_2789708.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_G10-L10-ep40/scores/scores_uncond_GOOG_G10-L10-ep40_integrated_2789708.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_G10-L10-ep40/scores/scores_uncond_GOOG_G10-L10-ep40_integrated_2789708.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.43085294591519613,
        "feature_points": {
          "spread": 0.8437160614174732,
          "ofi_up": 0.23322785179262337,
          "ask_cancellation_levels": 0.3696794505818462,
          "orderbook_imbalance": 0.15093389383326822,
          "bid_cancellation_levels": 0.35678542482186526,
          "ask_cancellation_depth": 0.7015497346390552,
          "limit_ask_order_levels": 0.32841555968097413,
          "limit_bid_order_levels": 0.34407319525620955,
          "vol_per_min": 0.1461833422322761,
          "ofi_down": 0.20220509432263079,
          "bid_volume_touch": 0.15903131454121958,
          "log_time_to_cancel": 0.18642654012511903,
          "limit_bid_order_depth": 0.6605217841796938,
          "ofi": 0.20491918889923333,
          "ask_volume_touch": 0.22402328819022765,
          "bid_cancellation_depth": 0.699108330191053,
          "limit_ask_order_depth": 0.7035798367742133,
          "ask_volume": 0.8190838369605807,
          "log_inter_arrival_time": 0.6755265385295199,
          "ofi_stay": 0.20543421464855996,
          "bid_volume": 0.8334873826014755
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_G10-L10-ep40/scores/scores_uncond_GOOG_G10-L10-ep40_integrated_2789708.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.430853,
            "report_record_id": "canonical-scores:37712c45-68fd-817b-b56d-ca3b4b014ce1",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-817c-88d6-cfa4b54c1597",
      "source_row_id": "37712c45-68fd-817c-88d6-cfa4b54c1597",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd817c88d6cfa4b54c1597",
      "run": "H2-55M-ctx4k-cond2000",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.136867,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.136867",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.114962,
      "reported_ks_context": 0.075894,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:32:56Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd817c88d6cfa4b54c1597",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:32:56Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:32:56Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:32:56Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.136867,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-817c-88d6-cfa4b54c1597",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond2000",
        "wws21_mean": 0.136867,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:32:56Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond2000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond2000_shard16_integrated_2617326.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond2000_GOOG_2617326/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:32:56Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond2000_GOOG_2617326/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd817c88d6cfa4b54c1597",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 842,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond2000_GOOG_2617326/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 2000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond2000_GOOG_2617326.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond2000_GOOG_2617326\"         --n_cond_msgs 2000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond2000_GOOG_2617326.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond2000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond2000_shard16_integrated_2617326.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond2000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond2000_integrated_2617326.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.13686718612005433,
        "feature_points": {
          "spread": 0.21509274274378565,
          "ask_cancellation_depth": 0.34780874526101146,
          "bid_cancellation_depth": 0.3323514709974648,
          "limit_ask_order_levels": 0.023733492737468338,
          "limit_bid_order_levels": 0.02844805009812171,
          "ask_cancellation_levels": 0.04331576885091604,
          "bid_cancellation_levels": 0.04086890028100026,
          "vol_per_min": 0.11210859599375556,
          "ofi": 0.05188806457912824,
          "ofi_up": 0.058421084577960844,
          "ofi_stay": 0.052981365968234065,
          "orderbook_imbalance": 0.08705896123263081,
          "ofi_down": 0.052606488746057854,
          "log_inter_arrival_time": 0.037756166009659725,
          "log_time_to_cancel": 0.36156025299214795,
          "ask_volume_touch": 0.07570431522955554,
          "bid_volume_touch": 0.1174828936774101,
          "ask_volume": 0.0815123663716619,
          "bid_volume": 0.12532807329118606,
          "limit_ask_order_depth": 0.3047942608239131,
          "limit_bid_order_depth": 0.3233888480580708
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond2000/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond2000_integrated_2617326.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.136867,
            "report_record_id": "canonical-scores:37712c45-68fd-817c-88d6-cfa4b54c1597",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-817c-9041-d85e0c4e5039",
      "source_row_id": "37712c45-68fd-817c-9041-d85e0c4e5039",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd817c9041d85e0c4e5039",
      "run": "adamw-32n-s26525",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.156173,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.156173",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.117851,
      "reported_ks_context": 0.080749,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:30Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd817c9041d85e0c4e5039",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:30Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:30Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:30Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.156173,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-817c-9041-d85e0c4e5039",
        "class": "G20",
        "model_evaluation": "adamw-32n-s26525",
        "wws21_mean": 0.156173,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:30Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s26525_GOOG_2871386/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s26525_GOOG_2871386/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:30Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd817c9041d85e0c4e5039",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 843,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s26525_GOOG_2871386.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 26525)",
            "checkpoint_step": 26525,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s26525_GOOG_2871386.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s26525_GOOG_2871386\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 26525         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s26525_GOOG_2871386.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s26525_GOOG_2871386\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 26525         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s26525_GOOG_2871386.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s26525_GOOG_2871386.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s26525_GOOG_2871386"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s26525_GOOG_2871386/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s26525_GOOG_2871386/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s26525_GOOG_2871386/data_real.tar",
            "selector": null,
            "source_value": 0.156173,
            "report_record_id": "canonical-scores:37712c45-68fd-817c-9041-d85e0c4e5039",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-817d-8295-e235224aef6a",
      "source_row_id": "37712c45-68fd-817d-8295-e235224aef6a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd817d8295e235224aef6a",
      "run": "muon-s80013",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.175479,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.175479",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.130384,
      "reported_ks_context": 0.09284,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:52Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd817d8295e235224aef6a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:52Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:52Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:52Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.175479,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-817d-8295-e235224aef6a",
        "class": "G20",
        "model_evaluation": "muon-s80013",
        "wws21_mean": 0.175479,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:52Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s80013_GOOG_2868874/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s80013_GOOG_2868874/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:52Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd817d8295e235224aef6a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 844,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s80013_GOOG_2868874.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248 (step 80013)",
            "checkpoint_step": 80013,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s80013_GOOG_2868874.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s80013_GOOG_2868874\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 80013         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263142+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.169342+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s80013_GOOG_2868874.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s80013_GOOG_2868874\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 80013         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s80013_GOOG_2868874.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s80013_GOOG_2868874.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-s80013_GOOG_2868874"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s80013_GOOG_2868874/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s80013_GOOG_2868874/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s80013_GOOG_2868874/data_real.tar",
            "selector": null,
            "source_value": 0.175479,
            "report_record_id": "canonical-scores:37712c45-68fd-817d-8295-e235224aef6a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-817d-82f0-f7465388b968",
      "source_row_id": "37712c45-68fd-817d-82f0-f7465388b968",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd817d82f0f7465388b968",
      "run": "genlen-ctx1k-c500-g1000",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.252783,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.252783",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.206666,
      "reported_ks_context": 0.142083,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:32Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd817d82f0f7465388b968",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:32Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:32Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:32Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.252783,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-817d-82f0-f7465388b968",
        "class": "G20",
        "model_evaluation": "genlen-ctx1k-c500-g1000",
        "wws21_mean": 0.252783,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:32Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1000_GOOG_3405778/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1000_GOOG_3405778/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:32Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd817d82f0f7465388b968",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 845,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1000_GOOG_3405778.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1000_GOOG_3405778.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx1k-c500-g1000_GOOG_3405778\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263039+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.145435+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 1000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1000_GOOG_3405778.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt1k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx1k-c500-g1000_GOOG_3405778\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1000_GOOG_3405778.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx1k-c500-g1000_GOOG_3405778.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx1k-c500-g1000_GOOG_3405778"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1000_GOOG_3405778/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1000_GOOG_3405778/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx1k-c500-g1000_GOOG_3405778/data_real.tar",
            "selector": null,
            "source_value": 0.252783,
            "report_record_id": "canonical-scores:37712c45-68fd-817d-82f0-f7465388b968",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-817e-bf2d-dae8d23598a5",
      "source_row_id": "37712c45-68fd-817e-bf2d-dae8d23598a5",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd817ebf2ddae8d23598a5",
      "run": "s4k-d512-tp4v12-step24530",
      "stock": null,
      "model_size": null,
      "run_family": "s4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.115443,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.115443",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.177279,
      "reported_ks_context": 0.102697,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:11Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd817ebf2ddae8d23598a5",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:11Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:11Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:11Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.115443,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-817e-bf2d-dae8d23598a5",
        "class": "G20",
        "model_evaluation": "s4k-d512-tp4v12-step24530",
        "wws21_mean": 0.115443,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:11Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-tp4v12-step24530_GOOG_3761914/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-tp4v12-step24530_GOOG_3761914/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:11Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd817ebf2ddae8d23598a5",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 846,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338/24530",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-tp4v12-step24530_GOOG_3761914.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338 (step 24530)",
            "checkpoint_step": 24530,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-tp4v12-step24530_GOOG_3761914.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-tp4v12-step24530_GOOG_3761914\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 24530         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.169438+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-tp4v12-step24530_GOOG_3761914.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-tp4v12-step24530_GOOG_3761914\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 24530         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-tp4v12-step24530_GOOG_3761914.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-tp4v12-step24530_GOOG_3761914.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s4k-d512-tp4v12-step24530_GOOG_3761914"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-tp4v12-step24530_GOOG_3761914/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-tp4v12-step24530_GOOG_3761914/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-tp4v12-step24530_GOOG_3761914/data_real.tar",
            "selector": null,
            "source_value": 0.115443,
            "report_record_id": "canonical-scores:37712c45-68fd-817e-bf2d-dae8d23598a5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-817f-808d-c43af1de37c7",
      "source_row_id": "37712c45-68fd-817f-808d-c43af1de37c7",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd817f808dc43af1de37c7",
      "run": "s4k-d512-step24530",
      "stock": null,
      "model_size": null,
      "run_family": "s4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.587635,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.587635",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.520134,
      "reported_ks_context": 0.455327,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:18Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd817f808dc43af1de37c7",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:18Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:18Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:18Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.587635,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-817f-808d-c43af1de37c7",
        "class": "G20",
        "model_evaluation": "s4k-d512-step24530",
        "wws21_mean": 0.587635,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:18Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-step24530_GOOG_3760423/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-step24530_GOOG_3760423/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:18Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd817f808dc43af1de37c7",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 847,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338/24530",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-step24530_GOOG_3760423.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338 (step 24530)",
            "checkpoint_step": 24530,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-step24530_GOOG_3760423.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-step24530_GOOG_3760423\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 24530         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.169438+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-step24530_GOOG_3760423.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3751338_dh0xciic_3751338\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-step24530_GOOG_3760423\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 24530         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-step24530_GOOG_3760423.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-step24530_GOOG_3760423.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s4k-d512-step24530_GOOG_3760423"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-step24530_GOOG_3760423/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-step24530_GOOG_3760423/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-step24530_GOOG_3760423/data_real.tar",
            "selector": null,
            "source_value": 0.587635,
            "report_record_id": "canonical-scores:37712c45-68fd-817f-808d-c43af1de37c7",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8180-90d3-f5f639fc7b64",
      "source_row_id": "37712c45-68fd-8180-90d3-f5f639fc7b64",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd818090d3f5f639fc7b64",
      "run": "pareto-120m-s91900",
      "stock": null,
      "model_size": "120M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.175299,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.175299",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.131017,
      "reported_ks_context": 0.097186,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:49Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd818090d3f5f639fc7b64",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:49Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:49Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:49Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.175299,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8180-90d3-f5f639fc7b64",
        "class": "G20",
        "model_evaluation": "pareto-120m-s91900",
        "wws21_mean": 0.175299,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:49Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s91900_GOOG_3253397/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s91900_GOOG_3253397/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:49Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd818090d3f5f639fc7b64",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 848,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292/91900",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s91900_GOOG_3253397.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292 (step 91900)",
            "checkpoint_step": 91900,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s91900_GOOG_3253397.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s91900_GOOG_3253397\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 91900         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.165112+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s91900_GOOG_3253397.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s91900_GOOG_3253397\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 91900         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s91900_GOOG_3253397.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s91900_GOOG_3253397.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s91900_GOOG_3253397"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s91900_GOOG_3253397/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s91900_GOOG_3253397/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s91900_GOOG_3253397/data_real.tar",
            "selector": null,
            "source_value": 0.175299,
            "report_record_id": "canonical-scores:37712c45-68fd-8180-90d3-f5f639fc7b64",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8180-982f-fd0d8dcb6da9",
      "source_row_id": "37712c45-68fd-8180-982f-fd0d8dcb6da9",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8180982ffd0d8dcb6da9",
      "run": "s4k-d256-s30080",
      "stock": null,
      "model_size": null,
      "run_family": "s4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.122775,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.122775",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.16006,
      "reported_ks_context": 0.101298,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:06Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8180982ffd0d8dcb6da9",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:06Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:06Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:06Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.122775,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8180-982f-fd0d8dcb6da9",
        "class": "G20",
        "model_evaluation": "s4k-d256-s30080",
        "wws21_mean": 0.122775,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:06Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s30080_GOOG_3848787/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s30080_GOOG_3848787/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:06Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8180982ffd0d8dcb6da9",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 849,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699/30080",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s30080_GOOG_3848787.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699 (step 30080)",
            "checkpoint_step": 30080,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s30080_GOOG_3848787.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d256-s30080_GOOG_3848787\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 30080         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.171015+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s30080_GOOG_3848787.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d256-s30080_GOOG_3848787\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 30080         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s30080_GOOG_3848787.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s30080_GOOG_3848787.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s4k-d256-s30080_GOOG_3848787"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s30080_GOOG_3848787/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s30080_GOOG_3848787/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s30080_GOOG_3848787/data_real.tar",
            "selector": null,
            "source_value": 0.122775,
            "report_record_id": "canonical-scores:37712c45-68fd-8180-982f-fd0d8dcb6da9",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8180-bfbf-d11cbd60e95e",
      "source_row_id": "37712c45-68fd-8180-bfbf-d11cbd60e95e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8180bfbfd11cbd60e95e",
      "run": "j2514440_step135458_24tok",
      "stock": null,
      "model_size": null,
      "run_family": "j2514440",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.134382,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.134382",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.117824,
      "reported_ks_context": 0.075862,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-01T08:19:43Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2150",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-01T08:19:43Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-01T08:19:43Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-01T08:19:43Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.134382,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8180-bfbf-d11cbd60e95e",
        "class": "G20",
        "model_evaluation": "j2514440_step135458_24tok",
        "wws21_mean": 0.134382,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-01T08:19:43Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440_step135458_24tok/scores/scores_uncond_GOOG_j2514440_step135458_24tok_shard4_integrated_2518009.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-01T08:19:43Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8180bfbfd11cbd60e95e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 850,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_GOOG_2518009.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 135458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_GOOG_2518009.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440_step135458_24tok/scores/scores_uncond_GOOG_j2514440_step135458_24tok_shard4_integrated_2518009.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440_step135458_24tok/scores/scores_uncond_GOOG_j2514440_step135458_24tok_integrated_2518009.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.134381929644474,
        "feature_points": {
          "spread": 0.17825713495974982,
          "ask_cancellation_depth": 0.3187137685929295,
          "bid_cancellation_depth": 0.325966649826433,
          "limit_ask_order_levels": 0.06287941660018229,
          "limit_bid_order_levels": 0.07466590898520135,
          "ask_cancellation_levels": 0.08212132026297445,
          "bid_cancellation_levels": 0.07156062942363556,
          "vol_per_min": 0.07259737054661505,
          "ofi": 0.04909453568875983,
          "ofi_up": 0.029500603148470632,
          "ofi_stay": 0.0503617331023333,
          "orderbook_imbalance": 0.11943404264139905,
          "ofi_down": 0.042019703752596636,
          "log_inter_arrival_time": 0.06570551752048691,
          "log_time_to_cancel": 0.3674359457736012,
          "ask_volume_touch": 0.06621965099994667,
          "bid_volume_touch": 0.05541910785409989,
          "ask_volume": 0.07013180473983316,
          "bid_volume": 0.047048056355463816,
          "limit_ask_order_depth": 0.3285509710442277,
          "limit_bid_order_depth": 0.3443366507150146
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440_step135458_24tok/scores/scores_uncond_GOOG_j2514440_step135458_24tok_integrated_2518009.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.134382,
            "report_record_id": "canonical-scores:37712c45-68fd-8180-bfbf-d11cbd60e95e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8181-9f07-f5ef880be0fc",
      "source_row_id": "37712c45-68fd-8181-9f07-f5ef880be0fc",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81819f07f5ef880be0fc",
      "run": "46m-s2k-d512",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.086913,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.086913",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.168934,
      "reported_ks_context": 0.095736,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:21Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81819f07f5ef880be0fc",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:21Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:21Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:21Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.086913,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8181-9f07-f5ef880be0fc",
        "class": "G20",
        "model_evaluation": "46m-s2k-d512",
        "wws21_mean": 0.086913,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:21Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d512_GOOG_3873789/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d512_GOOG_3873789/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:21Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81819f07f5ef880be0fc",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 851,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865272_n9fwqrjl_3865272/21830",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d512_GOOG_3873789.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865272_n9fwqrjl_3865272 (step 21830)",
            "checkpoint_step": 21830,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865272_n9fwqrjl_3865272",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d512_GOOG_3873789.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865272_n9fwqrjl_3865272\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d512_GOOG_3873789\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 21830         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.173305+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865272_n9fwqrjl_3865272",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865272_n9fwqrjl_3865272"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d512_GOOG_3873789.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865272_n9fwqrjl_3865272\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d512_GOOG_3873789\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 21830         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d512_GOOG_3873789.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d512_GOOG_3873789.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d512_GOOG_3873789"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d512_GOOG_3873789/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d512_GOOG_3873789/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d512_GOOG_3873789/data_real.tar",
            "selector": null,
            "source_value": 0.086913,
            "report_record_id": "canonical-scores:37712c45-68fd-8181-9f07-f5ef880be0fc",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8182-af33-d26f50cca237",
      "source_row_id": "37712c45-68fd-8182-af33-d26f50cca237",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8182af33d26f50cca237",
      "run": "pareto-55m-s29130",
      "stock": null,
      "model_size": "55M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.237201,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.237201",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.172519,
      "reported_ks_context": 0.128336,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:09Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8182af33d26f50cca237",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:09Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:09Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:09Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.237201,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8182-af33-d26f50cca237",
        "class": "G20",
        "model_evaluation": "pareto-55m-s29130",
        "wws21_mean": 0.237201,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:09Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s29130_GOOG_3253389/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s29130_GOOG_3253389/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:09Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8182af33d26f50cca237",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 852,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099/29130",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s29130_GOOG_3253389.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099 (step 29130)",
            "checkpoint_step": 29130,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s29130_GOOG_3253389.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s29130_GOOG_3253389\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 29130         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.132988+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s29130_GOOG_3253389.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s29130_GOOG_3253389\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 29130         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s29130_GOOG_3253389.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s29130_GOOG_3253389.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s29130_GOOG_3253389"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s29130_GOOG_3253389/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s29130_GOOG_3253389/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s29130_GOOG_3253389/data_real.tar",
            "selector": null,
            "source_value": 0.237201,
            "report_record_id": "canonical-scores:37712c45-68fd-8182-af33-d26f50cca237",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8183-bc6a-df66b8812696",
      "source_row_id": "37712c45-68fd-8183-bc6a-df66b8812696",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8183bc6adf66b8812696",
      "run": "j2504167-step100378-L100-c3500",
      "stock": null,
      "model_size": null,
      "run_family": "j2504167",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.146874,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.146874",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.12028,
      "reported_ks_context": 0.084391,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T21:58:46Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2180",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T21:58:46Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T21:58:46Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T21:58:46Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.146874,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8183-bc6a-df66b8812696",
        "class": "G20",
        "model_evaluation": "j2504167-step100378-L100-c3500",
        "wws21_mean": 0.146874,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T21:58:46Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167-step100378-L100-c3500/scores/scores_uncond_GOOG_j2504167-step100378-L100-c3500_integrated_2757355.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167-step100378-L100-c3500_GOOG_2757355/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T21:58:46Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167-step100378-L100-c3500_GOOG_2757355/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8183bc6adf66b8812696",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 853,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167-step100378-L100-c3500_GOOG_2757355/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=100378 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 3500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504167-step100378-L100-c3500_GOOG_2757355.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504167-step100378-L100-c3500_GOOG_2757355\"         --n_cond_msgs 3500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 100378         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504167-step100378-L100-c3500_GOOG_2757355.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167-step100378-L100-c3500/scores/scores_uncond_GOOG_j2504167-step100378-L100-c3500_integrated_2757355.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167-step100378-L100-c3500/scores/scores_uncond_GOOG_j2504167-step100378-L100-c3500_integrated_2757355.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1468741656609215,
        "feature_points": {
          "limit_bid_order_levels": 0.022294910160974967,
          "bid_cancellation_levels": 0.025317115822440988,
          "ofi_down": 0.08584455003478982,
          "bid_cancellation_depth": 0.32301935269453846,
          "spread": 0.21104300535382814,
          "ask_cancellation_levels": 0.039375209145755985,
          "ask_cancellation_depth": 0.2930609180073855,
          "log_inter_arrival_time": 0.035008476357648144,
          "log_time_to_cancel": 0.3423231584904566,
          "limit_ask_order_levels": 0.021074254481718735,
          "bid_volume": 0.2971319924222816,
          "vol_per_min": 0.06575392001269202,
          "ofi_up": 0.10779356137891427,
          "limit_bid_order_depth": 0.32871316462830247,
          "limit_ask_order_depth": 0.25879911361502445,
          "ask_volume_touch": 0.10240450079668817,
          "orderbook_imbalance": 0.09757435537320928,
          "bid_volume_touch": 0.08022770564215727,
          "ask_volume": 0.27846957158883434,
          "ofi": 0.035145415536727946,
          "ofi_stay": 0.0339832273349823
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167-step100378-L100-c3500/scores/scores_uncond_GOOG_j2504167-step100378-L100-c3500_integrated_2757355.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.146874,
            "report_record_id": "canonical-scores:37712c45-68fd-8183-bc6a-df66b8812696",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8184-bf35-e27bfd752e67",
      "source_row_id": "37712c45-68fd-8184-bf35-e27bfd752e67",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8184bf35e27bfd752e67",
      "run": "G10-l3k-bench",
      "stock": null,
      "model_size": null,
      "run_family": "G10",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 1.161962,
      "source_value_field": "Wasserstein",
      "source_value_text": "1.161962",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.868389,
      "reported_ks_context": 0.77526,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:57Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8184bf35e27bfd752e67",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:57Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:57Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:57Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 1.161962,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8184-bf35-e27bfd752e67",
        "class": "G20",
        "model_evaluation": "G10-l3k-bench",
        "wws21_mean": 1.161962,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:57Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-bench_GOOG_2868246/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-bench_GOOG_2868246/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:57Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8184bf35e27bfd752e67",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 854,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169/132203",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-bench_GOOG_2868246.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169 (step 132203)",
            "checkpoint_step": 132203,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-bench_GOOG_2868246.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l3k-bench_GOOG_2868246\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 132203         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.131932+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-bench_GOOG_2868246.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823169_0ktb7k5u_2823169\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l3k-bench_GOOG_2868246\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 132203         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-bench_GOOG_2868246.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-bench_GOOG_2868246.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l3k-bench_GOOG_2868246"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-bench_GOOG_2868246/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-bench_GOOG_2868246/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-bench_GOOG_2868246/data_real.tar",
            "selector": null,
            "source_value": 1.161962,
            "report_record_id": "canonical-scores:37712c45-68fd-8184-bf35-e27bfd752e67",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8186-8956-e88815805978",
      "source_row_id": "37712c45-68fd-8186-8956-e88815805978",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81868956e88815805978",
      "run": "N1-base-s63328-L100sim-fix",
      "stock": null,
      "model_size": null,
      "run_family": "N1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.227526,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.227526",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.186297,
      "reported_ks_context": 0.13901,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:44Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81868956e88815805978",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:44Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:44Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:44Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.227526,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8186-8956-e88815805978",
        "class": "G20",
        "model_evaluation": "N1-base-s63328-L100sim-fix",
        "wws21_mean": 0.227526,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:44Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim-fix_GOOG_3234779/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim-fix_GOOG_3234779/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:44Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81868956e88815805978",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 855,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391/63328",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim-fix_GOOG_3234779.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391 (step 63328)",
            "checkpoint_step": 63328,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim-fix_GOOG_3234779.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1-base-s63328-L100sim-fix_GOOG_3234779\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 63328         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.143072+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim-fix_GOOG_3234779.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1-base-s63328-L100sim-fix_GOOG_3234779\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 63328         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim-fix_GOOG_3234779.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100sim-fix_GOOG_3234779.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1-base-s63328-L100sim-fix_GOOG_3234779"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim-fix_GOOG_3234779/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim-fix_GOOG_3234779/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100sim-fix_GOOG_3234779/data_real.tar",
            "selector": null,
            "source_value": 0.227526,
            "report_record_id": "canonical-scores:37712c45-68fd-8186-8956-e88815805978",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8186-9a91-ce4002d4918e",
      "source_row_id": "37712c45-68fd-8186-9a91-ce4002d4918e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81869a91ce4002d4918e",
      "run": "s2k-d256-s55040",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.097627,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.097627",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.145305,
      "reported_ks_context": 0.08816,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:34Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81869a91ce4002d4918e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:34Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:34Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:34Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.097627,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8186-9a91-ce4002d4918e",
        "class": "G20",
        "model_evaluation": "s2k-d256-s55040",
        "wws21_mean": 0.097627,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:34Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s55040_GOOG_3835485/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s55040_GOOG_3835485/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:34Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81869a91ce4002d4918e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 856,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596/55040",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s55040_GOOG_3835485.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596 (step 55040)",
            "checkpoint_step": 55040,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s55040_GOOG_3835485.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s55040_GOOG_3835485\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 55040         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.121699+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s55040_GOOG_3835485.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s55040_GOOG_3835485\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 55040         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s55040_GOOG_3835485.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s55040_GOOG_3835485.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-d256-s55040_GOOG_3835485"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s55040_GOOG_3835485/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s55040_GOOG_3835485/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s55040_GOOG_3835485/data_real.tar",
            "selector": null,
            "source_value": 0.097627,
            "report_record_id": "canonical-scores:37712c45-68fd-8186-9a91-ce4002d4918e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8187-8089-c1174e21bbdf",
      "source_row_id": "37712c45-68fd-8187-8089-c1174e21bbdf",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81878089c1174e21bbdf",
      "run": "46m-s2k-d256-nobook-hier-c1000g250",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.120127,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.120127",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.161256,
      "reported_ks_context": 0.103228,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:54Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81878089c1174e21bbdf",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:54Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:54Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:54Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.120127,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8187-8089-c1174e21bbdf",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-nobook-hier-c1000g250",
        "wws21_mean": 0.120127,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:54Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:54Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81878089c1174e21bbdf",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 857,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301/37610",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301 (step 37610)",
            "checkpoint_step": 37610,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047\"         --n_cond_msgs 1000         --n_gen_msgs 250         --n_sequences 1024         --batch_size 32         --checkpoint_step 37610         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.174479+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 1000,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047\"         --n_cond_msgs 1000         --n_gen_msgs 250         --n_sequences 1024         --batch_size 32         --checkpoint_step 37610         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1000g250_GOOG_3876047/data_real.tar",
            "selector": null,
            "source_value": 0.120127,
            "report_record_id": "canonical-scores:37712c45-68fd-8187-8089-c1174e21bbdf",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8187-b57c-eeba89f569e0",
      "source_row_id": "37712c45-68fd-8187-b57c-eeba89f569e0",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8187b57ceeba89f569e0",
      "run": "adamw-32n-s24615",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.147231,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.147231",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.111269,
      "reported_ks_context": 0.077995,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:54Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8187b57ceeba89f569e0",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:54Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:54Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:54Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.147231,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8187-b57c-eeba89f569e0",
        "class": "G20",
        "model_evaluation": "adamw-32n-s24615",
        "wws21_mean": 0.147231,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:54Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s24615_GOOG_2871385/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s24615_GOOG_2871385/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:54Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8187b57ceeba89f569e0",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 858,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s24615_GOOG_2871385.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 24615)",
            "checkpoint_step": 24615,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s24615_GOOG_2871385.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s24615_GOOG_2871385\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 24615         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s24615_GOOG_2871385.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s24615_GOOG_2871385\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 24615         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s24615_GOOG_2871385.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s24615_GOOG_2871385.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s24615_GOOG_2871385"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s24615_GOOG_2871385/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s24615_GOOG_2871385/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s24615_GOOG_2871385/data_real.tar",
            "selector": null,
            "source_value": 0.147231,
            "report_record_id": "canonical-scores:37712c45-68fd-8187-b57c-eeba89f569e0",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-818a-b247-f5020bc5e980",
      "source_row_id": "37712c45-68fd-818a-b247-f5020bc5e980",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd818ab247f5020bc5e980",
      "run": "R1-s4k-d256-raw-s53730",
      "stock": null,
      "model_size": null,
      "run_family": "R1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.171865,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.171865",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.170908,
      "reported_ks_context": 0.100567,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:40Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd818ab247f5020bc5e980",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:40Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:40Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:40Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.171865,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-818a-b247-f5020bc5e980",
        "class": "G20",
        "model_evaluation": "R1-s4k-d256-raw-s53730",
        "wws21_mean": 0.171865,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:40Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s53730_GOOG_3745433/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s53730_GOOG_3745433/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:40Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd818ab247f5020bc5e980",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 859,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292/53730",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s53730_GOOG_3745433.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292 (step 53730)",
            "checkpoint_step": 53730,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s53730_GOOG_3745433.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s53730_GOOG_3745433\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 53730         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.163252+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s53730_GOOG_3745433.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3744292_z65jz92d_3744292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s53730_GOOG_3745433\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 53730         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s53730_GOOG_3745433.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s53730_GOOG_3745433.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s53730_GOOG_3745433"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s53730_GOOG_3745433/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s53730_GOOG_3745433/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s53730_GOOG_3745433/data_real.tar",
            "selector": null,
            "source_value": 0.171865,
            "report_record_id": "canonical-scores:37712c45-68fd-818a-b247-f5020bc5e980",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-818d-858f-cec79a7fc9dd",
      "source_row_id": "37712c45-68fd-818d-858f-cec79a7fc9dd",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd818d858fcec79a7fc9dd",
      "run": "s4k-d512-32n-step14980",
      "stock": null,
      "model_size": null,
      "run_family": "s4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.138068,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.138068",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.18125,
      "reported_ks_context": 0.120368,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd818d858fcec79a7fc9dd",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.138068,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-818d-858f-cec79a7fc9dd",
        "class": "G20",
        "model_evaluation": "s4k-d512-32n-step14980",
        "wws21_mean": 0.138068,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step14980_GOOG_3762113/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step14980_GOOG_3762113/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd818d858fcec79a7fc9dd",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 860,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036/14980",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step14980_GOOG_3762113.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036 (step 14980)",
            "checkpoint_step": 14980,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step14980_GOOG_3762113.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step14980_GOOG_3762113\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 14980         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.176082+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step14980_GOOG_3762113.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step14980_GOOG_3762113\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 14980         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step14980_GOOG_3762113.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step14980_GOOG_3762113.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step14980_GOOG_3762113"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step14980_GOOG_3762113/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step14980_GOOG_3762113/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step14980_GOOG_3762113/data_real.tar",
            "selector": null,
            "source_value": 0.138068,
            "report_record_id": "canonical-scores:37712c45-68fd-818d-858f-cec79a7fc9dd",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-818d-882f-c6e3727a1dd2",
      "source_row_id": "37712c45-68fd-818d-882f-c6e3727a1dd2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd818d882fc6e3727a1dd2",
      "run": "adamw-s49871",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.182778,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.182778",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.142358,
      "reported_ks_context": 0.09514,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:59Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd818d882fc6e3727a1dd2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:59Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:59Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:59Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.182778,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-818d-882f-c6e3727a1dd2",
        "class": "G20",
        "model_evaluation": "adamw-s49871",
        "wws21_mean": 0.182778,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:59Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s49871_GOOG_2868485/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s49871_GOOG_2868485/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:59Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd818d882fc6e3727a1dd2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 861,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s49871_GOOG_2868485.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665 (step 49871)",
            "checkpoint_step": 49871,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s49871_GOOG_2868485.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s49871_GOOG_2868485\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 49871         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262928+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.121688+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s49871_GOOG_2868485.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s49871_GOOG_2868485\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 49871         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s49871_GOOG_2868485.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s49871_GOOG_2868485.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-s49871_GOOG_2868485"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s49871_GOOG_2868485/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s49871_GOOG_2868485/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s49871_GOOG_2868485/data_real.tar",
            "selector": null,
            "source_value": 0.182778,
            "report_record_id": "canonical-scores:37712c45-68fd-818d-882f-c6e3727a1dd2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-818e-9c7b-dde45e1e7aab",
      "source_row_id": "37712c45-68fd-818e-9c7b-dde45e1e7aab",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd818e9c7bdde45e1e7aab",
      "run": "46m-s2k-d256-nobook-hier",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.10978,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.10978",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.168298,
      "reported_ks_context": 0.10775,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:38Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd818e9c7bdde45e1e7aab",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:38Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:38Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:38Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.10978,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-818e-9c7b-dde45e1e7aab",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-nobook-hier",
        "wws21_mean": 0.10978,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:38Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier_GOOG_3874901/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier_GOOG_3874901/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:38Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd818e9c7bdde45e1e7aab",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 862,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301/37610",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier_GOOG_3874901.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301 (step 37610)",
            "checkpoint_step": 37610,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier_GOOG_3874901.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier_GOOG_3874901\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 37610         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.174479+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier_GOOG_3874901.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier_GOOG_3874901\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 37610         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier_GOOG_3874901.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier_GOOG_3874901.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier_GOOG_3874901"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier_GOOG_3874901/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier_GOOG_3874901/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier_GOOG_3874901/data_real.tar",
            "selector": null,
            "source_value": 0.10978,
            "report_record_id": "canonical-scores:37712c45-68fd-818e-9c7b-dde45e1e7aab",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8190-a286-e91fbb311b91",
      "source_row_id": "37712c45-68fd-8190-a286-e91fbb311b91",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8190a286e91fbb311b91",
      "run": "s2k-d256-s75740",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.121038,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.121038",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.162579,
      "reported_ks_context": 0.090613,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:10Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8190a286e91fbb311b91",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:10Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:10Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:10Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.121038,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8190-a286-e91fbb311b91",
        "class": "G20",
        "model_evaluation": "s2k-d256-s75740",
        "wws21_mean": 0.121038,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:10Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s75740_GOOG_3848657/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s75740_GOOG_3848657/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:10Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8190a286e91fbb311b91",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 863,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596/75740",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s75740_GOOG_3848657.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596 (step 75740)",
            "checkpoint_step": 75740,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s75740_GOOG_3848657.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s75740_GOOG_3848657\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 75740         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:13:15.378012+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s75740_GOOG_3848657.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s75740_GOOG_3848657\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 75740         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s75740_GOOG_3848657.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s75740_GOOG_3848657.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-d256-s75740_GOOG_3848657"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s75740_GOOG_3848657/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s75740_GOOG_3848657/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s75740_GOOG_3848657/data_real.tar",
            "selector": null,
            "source_value": 0.121038,
            "report_record_id": "canonical-scores:37712c45-68fd-8190-a286-e91fbb311b91",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8190-a466-ef776d7d05cb",
      "source_row_id": "37712c45-68fd-8190-a466-ef776d7d05cb",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8190a466ef776d7d05cb",
      "run": "H2-55M-ctx8k-cond7500",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.15791,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.15791",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.117262,
      "reported_ks_context": 0.083932,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:30Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8190a466ef776d7d05cb",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:30Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:30Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:30Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.15791,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8190-a466-ef776d7d05cb",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx8k-cond7500",
        "wws21_mean": 0.15791,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:30Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-cond7500/scores/scores_uncond_GOOG_H2-55M-ctx8k-cond7500_shard12_integrated_2597837.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-cond7500_GOOG_2597837/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:30Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-cond7500_GOOG_2597837/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8190a466ef776d7d05cb",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 864,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "58091",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-cond7500_GOOG_2597837/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "41850",
                "45450",
                "46650",
                "47850",
                "49050",
                "50250",
                "51450",
                "52650",
                "53849",
                "55048",
                "56248",
                "57434",
                "58091",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608/58091",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 7500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx8k-cond7500_GOOG_2597837.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-8k/checkpoints/j2506608_7nd1n750_2506608\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx8k-cond7500_GOOG_2597837\"         --n_cond_msgs 7500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 4         --checkpoint_step 58091         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx8k-cond7500_GOOG_2597837.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-cond7500/scores/scores_uncond_GOOG_H2-55M-ctx8k-cond7500_shard12_integrated_2597837.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-cond7500/scores/scores_uncond_GOOG_H2-55M-ctx8k-cond7500_integrated_2597837.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15790986601375984,
        "feature_points": {
          "spread": 0.3239634245870799,
          "ask_cancellation_depth": 0.3199552607053772,
          "bid_cancellation_depth": 0.30763837229416474,
          "limit_ask_order_levels": 0.020344565098857723,
          "limit_bid_order_levels": 0.014422701020461316,
          "ask_cancellation_levels": 0.06179105796032349,
          "bid_cancellation_levels": 0.04470848070318161,
          "vol_per_min": 0.15683295536641856,
          "ofi": 0.0428072654228232,
          "ofi_up": 0.0820672907068042,
          "ofi_stay": 0.04245725760452299,
          "orderbook_imbalance": 0.10470536678146687,
          "ofi_down": 0.06982447003200937,
          "log_inter_arrival_time": 0.03799601026899848,
          "log_time_to_cancel": 0.3651206290492496,
          "ask_volume_touch": 0.16338918724294849,
          "bid_volume_touch": 0.12538866762487272,
          "ask_volume": 0.2177971862569645,
          "bid_volume": 0.19874503498786344,
          "limit_ask_order_depth": 0.31442076761987103,
          "limit_bid_order_depth": 0.301731234954697
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx8k-cond7500/scores/scores_uncond_GOOG_H2-55M-ctx8k-cond7500_integrated_2597837.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.15791,
            "report_record_id": "canonical-scores:37712c45-68fd-8190-a466-ef776d7d05cb",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8192-80cf-ee05a249cffe",
      "source_row_id": "37712c45-68fd-8192-80cf-ee05a249cffe",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd819280cfee05a249cffe",
      "run": "ctx2k-cond500",
      "stock": null,
      "model_size": null,
      "run_family": "ctx2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.144095,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.144095",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.12145,
      "reported_ks_context": 0.085796,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:05Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd819280cfee05a249cffe",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:05Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:05Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:05Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.144095,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8192-80cf-ee05a249cffe",
        "class": "G20",
        "model_evaluation": "ctx2k-cond500",
        "wws21_mean": 0.144095,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:05Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond500/scores/scores_uncond_GOOG_ctx2k-cond500_integrated_2826366.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond500_GOOG_2826366/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:05Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond500_GOOG_2826366/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd819280cfee05a249cffe",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 865,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "116460",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond500_GOOG_2826366/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=116460 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950/116460",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "101180",
                "103865",
                "106554",
                "109240",
                "111923",
                "114608",
                "116460",
                "60915",
                "87740",
                "90425",
                "93115",
                "95804",
                "98489",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950/116460",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx2k-cond500_GOOG_2826366.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx2k-cond500_GOOG_2826366\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 116460         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx2k-cond500_GOOG_2826366.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond500/scores/scores_uncond_GOOG_ctx2k-cond500_integrated_2826366.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond500/scores/scores_uncond_GOOG_ctx2k-cond500_integrated_2826366.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14409506072501038,
        "feature_points": {
          "spread": 0.29590710476619003,
          "bid_cancellation_depth": 0.3123338631844834,
          "ofi_down": 0.042195286391693634,
          "ask_volume": 0.2141583905075669,
          "limit_ask_order_depth": 0.31047808650386644,
          "ask_cancellation_levels": 0.0418566154336231,
          "limit_bid_order_levels": 0.028616584271359387,
          "ofi_up": 0.05004485998072813,
          "vol_per_min": 0.04949585413057568,
          "bid_cancellation_levels": 0.04184801171143696,
          "limit_ask_order_levels": 0.024373778119406796,
          "ask_cancellation_depth": 0.29917409948712803,
          "limit_bid_order_depth": 0.32571244833819435,
          "orderbook_imbalance": 0.10715740087771702,
          "ask_volume_touch": 0.07523745387605582,
          "bid_volume": 0.2342699869450841,
          "ofi": 0.04761703272769379,
          "ofi_stay": 0.04790829780965233,
          "log_time_to_cancel": 0.39492918629133517,
          "bid_volume_touch": 0.06511483867790241,
          "log_inter_arrival_time": 0.017567095193524247
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond500/scores/scores_uncond_GOOG_ctx2k-cond500_integrated_2826366.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.144095,
            "report_record_id": "canonical-scores:37712c45-68fd-8192-80cf-ee05a249cffe",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8193-9476-c033440210f5",
      "source_row_id": "37712c45-68fd-8193-9476-c033440210f5",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81939476c033440210f5",
      "run": "s4k-d512-32n-step24520",
      "stock": null,
      "model_size": null,
      "run_family": "s4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.106022,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.106022",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.163348,
      "reported_ks_context": 0.098447,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:03Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81939476c033440210f5",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:03Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:03Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:03Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.106022,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8193-9476-c033440210f5",
        "class": "G20",
        "model_evaluation": "s4k-d512-32n-step24520",
        "wws21_mean": 0.106022,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:03Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step24520_GOOG_3763548/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step24520_GOOG_3763548/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:03Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 868 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81939476c033440210f5",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 866,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036/24520",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step24520_GOOG_3763548.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036 (step 24520)",
            "checkpoint_step": 24520,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step24520_GOOG_3763548.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step24520_GOOG_3763548\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 24520         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.176082+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step24520_GOOG_3763548.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step24520_GOOG_3763548\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 24520         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step24520_GOOG_3763548.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step24520_GOOG_3763548.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step24520_GOOG_3763548"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step24520_GOOG_3763548/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step24520_GOOG_3763548/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step24520_GOOG_3763548/data_real.tar",
            "selector": null,
            "source_value": 0.106022,
            "report_record_id": "canonical-scores:37712c45-68fd-8193-9476-c033440210f5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8193-98ef-f39735d2efee",
      "source_row_id": "37712c45-68fd-8193-98ef-f39735d2efee",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd819398eff39735d2efee",
      "run": "h200_23tok_step37k",
      "stock": null,
      "model_size": null,
      "run_family": "h200",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.188642,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.188642",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.151472,
      "reported_ks_context": 0.100422,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:38Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd819398eff39735d2efee",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:38Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:38Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:38Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.188642,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8193-98ef-f39735d2efee",
        "class": "G20",
        "model_evaluation": "h200_23tok_step37k",
        "wws21_mean": 0.188642,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:38Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step37k/scores/scores_uncond_GOOG_h200_23tok_step37k_shard20_integrated_2570837.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/h200_23tok_step37k_GOOG_2570837/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:38Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/h200_23tok_step37k_GOOG_2570837/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd819398eff39735d2efee",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 867,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "37455",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/37455",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/h200_23tok_step37k_GOOG_2570837/log_rank0.txt",
            "evidence_line": 628,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/37455/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/37455",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/37455",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "103505",
                "108205",
                "112915",
                "160000",
                "164710",
                "174140",
                "178853",
                "183557",
                "188270",
                "192987",
                "197705",
                "202424",
                "207134",
                "211843",
                "216558",
                "23310",
                "28025",
                "32740",
                "37455",
                "56320",
                "70465",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/37455",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step37k_GOOG_2570837.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/LOBS5/inference_results/h200_23tok_step37k_GOOG_2570837\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 37455         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_h200_23tok_step37k_GOOG_2570837.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step37k/scores/scores_uncond_GOOG_h200_23tok_step37k_shard20_integrated_2570837.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step37k/scores/scores_uncond_GOOG_h200_23tok_step37k_integrated_2570837.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.18864155468494131,
        "feature_points": {
          "spread": 0.4441060122139944,
          "ask_cancellation_depth": 0.36055720751831455,
          "bid_cancellation_depth": 0.3427751315590428,
          "limit_ask_order_levels": 0.09528069055018432,
          "limit_bid_order_levels": 0.08199174890032265,
          "ask_cancellation_levels": 0.14051280791895682,
          "bid_cancellation_levels": 0.15559750791380225,
          "vol_per_min": 0.1797189147187965,
          "ofi": 0.06811578248021116,
          "ofi_up": 0.09797843609855447,
          "ofi_stay": 0.06919000595586865,
          "orderbook_imbalance": 0.12222395855178121,
          "ofi_down": 0.07243019449488311,
          "log_inter_arrival_time": 0.05057601363746768,
          "log_time_to_cancel": 0.5259151278529504,
          "ask_volume_touch": 0.10086863371759491,
          "bid_volume_touch": 0.09117623048374524,
          "ask_volume": 0.11393952204237248,
          "bid_volume": 0.12000086539412114,
          "limit_ask_order_depth": 0.37480683653955227,
          "limit_bid_order_depth": 0.3537110198412506
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_h200_23tok_step37k/scores/scores_uncond_GOOG_h200_23tok_step37k_integrated_2570837.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.188642,
            "report_record_id": "canonical-scores:37712c45-68fd-8193-98ef-f39735d2efee",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8198-9020-e700a451caca",
      "source_row_id": "37712c45-68fd-8198-9020-e700a451caca",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81989020e700a451caca",
      "run": "nsa-o2b-FINAL",
      "stock": null,
      "model_size": null,
      "run_family": "nsa",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.929049,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.929049",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.685994,
      "reported_ks_context": 0.580771,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:42Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81989020e700a451caca",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:42Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:42Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:42Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.929049,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8198-9020-e700a451caca",
        "class": "G20",
        "model_evaluation": "nsa-o2b-FINAL",
        "wws21_mean": 0.929049,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:42Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-FINAL_GOOG_3256270/data_real",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-FINAL_GOOG_3256270/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:42Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81989020e700a451caca",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 868,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922/14794",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-FINAL_GOOG_3256270.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922 (step 14794)",
            "checkpoint_step": 14794,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-FINAL_GOOG_3256270.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-FINAL_GOOG_3256270\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.148778+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-FINAL_GOOG_3256270.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-FINAL_GOOG_3256270\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-FINAL_GOOG_3256270.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-FINAL_GOOG_3256270.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/nsa-o2b-FINAL_GOOG_3256270"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-FINAL_GOOG_3256270/data_real",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-FINAL_GOOG_3256270/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-FINAL_GOOG_3256270/data_real",
            "selector": null,
            "source_value": 0.929049,
            "report_record_id": "canonical-scores:37712c45-68fd-8198-9020-e700a451caca",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8198-abf9-e91a8e1e6580",
      "source_row_id": "37712c45-68fd-8198-abf9-e91a8e1e6580",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8198abf9e91a8e1e6580",
      "run": "s2k-cooldown-s83310",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.082425,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.082425",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.136334,
      "reported_ks_context": 0.083031,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:27Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8198abf9e91a8e1e6580",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:27Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:27Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:27Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.082425,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8198-abf9-e91a8e1e6580",
        "class": "G20",
        "model_evaluation": "s2k-cooldown-s83310",
        "wws21_mean": 0.082425,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:27Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s83310_GOOG_3855692/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s83310_GOOG_3855692/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:27Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8198abf9e91a8e1e6580",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 869,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/83310",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s83310_GOOG_3855692.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 83310)",
            "checkpoint_step": 83310,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s83310_GOOG_3855692.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-cooldown-s83310_GOOG_3855692\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 83310         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s83310_GOOG_3855692.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-cooldown-s83310_GOOG_3855692\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 83310         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s83310_GOOG_3855692.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s83310_GOOG_3855692.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-cooldown-s83310_GOOG_3855692"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s83310_GOOG_3855692/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s83310_GOOG_3855692/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s83310_GOOG_3855692/data_real.tar",
            "selector": null,
            "source_value": 0.082425,
            "report_record_id": "canonical-scores:37712c45-68fd-8198-abf9-e91a8e1e6580",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8198-bfa0-ea8625be3870",
      "source_row_id": "37712c45-68fd-8198-bfa0-ea8625be3870",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd8198bfa0ea8625be3870",
      "run": "R1-s4k-d256-step13020",
      "stock": null,
      "model_size": null,
      "run_family": "R1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.133995,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.133995",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.194133,
      "reported_ks_context": 0.115835,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:34Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd8198bfa0ea8625be3870",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:34Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:34Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:34Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.133995,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8198-bfa0-ea8625be3870",
        "class": "G20",
        "model_evaluation": "R1-s4k-d256-step13020",
        "wws21_mean": 0.133995,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:34Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step13020_GOOG_3742892/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step13020_GOOG_3742892/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:34Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd8198bfa0ea8625be3870",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 870,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133/13020",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step13020_GOOG_3742892.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133 (step 13020)",
            "checkpoint_step": 13020,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step13020_GOOG_3742892.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-step13020_GOOG_3742892\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 13020         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.146737+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step13020_GOOG_3742892.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3728133_ei9u39ly_3728133\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-step13020_GOOG_3742892\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 13020         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step13020_GOOG_3742892.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-step13020_GOOG_3742892.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-step13020_GOOG_3742892"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step13020_GOOG_3742892/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step13020_GOOG_3742892/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-step13020_GOOG_3742892/data_real.tar",
            "selector": null,
            "source_value": 0.133995,
            "report_record_id": "canonical-scores:37712c45-68fd-8198-bfa0-ea8625be3870",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-8199-9100-fe02e494e950",
      "source_row_id": "37712c45-68fd-8199-9100-fe02e494e950",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81999100fe02e494e950",
      "run": "46m-s2k-d256-bws0",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.15588,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.15588",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.178406,
      "reported_ks_context": 0.120954,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:45Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81999100fe02e494e950",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:45Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:45Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:45Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.15588,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-8199-9100-fe02e494e950",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-bws0",
        "wws21_mean": 0.15588,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:45Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-bws0_GOOG_3864530/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-bws0_GOOG_3864530/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:45Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81999100fe02e494e950",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 871,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/83310",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-bws0_GOOG_3864530.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 83310)",
            "checkpoint_step": 83310,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-bws0_GOOG_3864530.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R11_book-ablation         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-bws0_GOOG_3864530\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 83310         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100 --book_weight_scale 0"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-bws0_GOOG_3864530.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R11_book-ablation         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-bws0_GOOG_3864530\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 83310         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100 --book_weight_scale 0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-bws0_GOOG_3864530.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-bws0_GOOG_3864530.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-bws0_GOOG_3864530"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-bws0_GOOG_3864530/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-bws0_GOOG_3864530/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-bws0_GOOG_3864530/data_real.tar",
            "selector": null,
            "source_value": 0.15588,
            "report_record_id": "canonical-scores:37712c45-68fd-8199-9100-fe02e494e950",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-819a-8209-c0f8ac74452c",
      "source_row_id": "37712c45-68fd-819a-8209-c0f8ac74452c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd819a8209c0f8ac74452c",
      "run": "26tok-muon-s67479-1024seq",
      "stock": null,
      "model_size": null,
      "run_family": "26tok",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.201966,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.201966",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.204158,
      "reported_ks_context": 0.14137,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:32:59Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd819a8209c0f8ac74452c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:32:59Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:32:59Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:32:59Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.201966,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-819a-8209-c0f8ac74452c",
        "class": "G20",
        "model_evaluation": "26tok-muon-s67479-1024seq",
        "wws21_mean": 0.201966,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:32:59Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/26tok-muon-s67479-1024seq_GOOG_3114851/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/26tok-muon-s67479-1024seq_GOOG_3114851/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:32:59Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd819a8209c0f8ac74452c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 872,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3108754_y4vbpmi3_3108754/67479",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_26tok-muon-s67479-1024seq_GOOG_3114851.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3108754_y4vbpmi3_3108754 (step 67479)",
            "checkpoint_step": 67479,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3108754_y4vbpmi3_3108754",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_26tok-muon-s67479-1024seq_GOOG_3114851.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3108754_y4vbpmi3_3108754\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/26tok-muon-s67479-1024seq_GOOG_3114851\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 67479         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.178695+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3108754_y4vbpmi3_3108754",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3108754_y4vbpmi3_3108754"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_26tok-muon-s67479-1024seq_GOOG_3114851.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3108754_y4vbpmi3_3108754\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/26tok-muon-s67479-1024seq_GOOG_3114851\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 67479         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_26tok-muon-s67479-1024seq_GOOG_3114851.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_26tok-muon-s67479-1024seq_GOOG_3114851.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/26tok-muon-s67479-1024seq_GOOG_3114851"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/26tok-muon-s67479-1024seq_GOOG_3114851/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/26tok-muon-s67479-1024seq_GOOG_3114851/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/26tok-muon-s67479-1024seq_GOOG_3114851/data_real.tar",
            "selector": null,
            "source_value": 0.201966,
            "report_record_id": "canonical-scores:37712c45-68fd-819a-8209-c0f8ac74452c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-819a-bbe1-d76bd55eac49",
      "source_row_id": "37712c45-68fd-819a-bbe1-d76bd55eac49",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd819abbe1d76bd55eac49",
      "run": "genlen-ctx4k-c500-g500",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.213086,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.213086",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.163877,
      "reported_ks_context": 0.110526,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:07Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd819abbe1d76bd55eac49",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:07Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:07Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:07Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.213086,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-819a-bbe1-d76bd55eac49",
        "class": "G20",
        "model_evaluation": "genlen-ctx4k-c500-g500",
        "wws21_mean": 0.213086,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:07Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g500_GOOG_3405784/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g500_GOOG_3405784/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:07Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd819abbe1d76bd55eac49",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 873,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g500_GOOG_3405784.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g500_GOOG_3405784.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g500_GOOG_3405784\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262982+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134305+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g500_GOOG_3405784.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g500_GOOG_3405784\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g500_GOOG_3405784.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g500_GOOG_3405784.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g500_GOOG_3405784"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g500_GOOG_3405784/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g500_GOOG_3405784/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g500_GOOG_3405784/data_real.tar",
            "selector": null,
            "source_value": 0.213086,
            "report_record_id": "canonical-scores:37712c45-68fd-819a-bbe1-d76bd55eac49",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-819b-b70d-d3776a8b3205",
      "source_row_id": "37712c45-68fd-819b-b70d-d3776a8b3205",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd819bb70dd3776a8b3205",
      "run": "pareto-360m-adamw32n-s24615",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.170894,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.170894",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.125569,
      "reported_ks_context": 0.092901,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:02Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd819bb70dd3776a8b3205",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:02Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:02Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:02Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.170894,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-819b-b70d-d3776a8b3205",
        "class": "G20",
        "model_evaluation": "pareto-360m-adamw32n-s24615",
        "wws21_mean": 0.170894,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:02Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s24615_GOOG_3253406/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s24615_GOOG_3253406/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:02Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd819bb70dd3776a8b3205",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 874,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s24615_GOOG_3253406.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 24615)",
            "checkpoint_step": 24615,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s24615_GOOG_3253406.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw32n-s24615_GOOG_3253406\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 24615         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s24615_GOOG_3253406.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw32n-s24615_GOOG_3253406\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 24615         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s24615_GOOG_3253406.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s24615_GOOG_3253406.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw32n-s24615_GOOG_3253406"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s24615_GOOG_3253406/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s24615_GOOG_3253406/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s24615_GOOG_3253406/data_real.tar",
            "selector": null,
            "source_value": 0.170894,
            "report_record_id": "canonical-scores:37712c45-68fd-819b-b70d-d3776a8b3205",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-819c-b1ff-c5e7fdce4c51",
      "source_row_id": "37712c45-68fd-819c-b1ff-c5e7fdce4c51",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd819cb1ffc5e7fdce4c51",
      "run": "adamw-s84249",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.153422,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.153422",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.118816,
      "reported_ks_context": 0.085468,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:13Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd819cb1ffc5e7fdce4c51",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:13Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:13Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:13Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.153422,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-819c-b1ff-c5e7fdce4c51",
        "class": "G20",
        "model_evaluation": "adamw-s84249",
        "wws21_mean": 0.153422,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:13Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s84249_GOOG_2868489/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s84249_GOOG_2868489/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:13Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd819cb1ffc5e7fdce4c51",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 875,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s84249_GOOG_2868489.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661 (step 84249)",
            "checkpoint_step": 84249,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s84249_GOOG_2868489.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s84249_GOOG_2868489\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 84249         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263172+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.179826+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s84249_GOOG_2868489.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s84249_GOOG_2868489\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 84249         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s84249_GOOG_2868489.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s84249_GOOG_2868489.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-s84249_GOOG_2868489"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s84249_GOOG_2868489/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s84249_GOOG_2868489/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s84249_GOOG_2868489/data_real.tar",
            "selector": null,
            "source_value": 0.153422,
            "report_record_id": "canonical-scores:37712c45-68fd-819c-b1ff-c5e7fdce4c51",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-819d-ad66-ebd03b6f65d4",
      "source_row_id": "37712c45-68fd-819d-ad66-ebd03b6f65d4",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd819dad66ebd03b6f65d4",
      "run": "14m-s2k-d256-c1000g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "14m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.098007,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.098007",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.148765,
      "reported_ks_context": 0.103845,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:09Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd819dad66ebd03b6f65d4",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:09Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:09Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:09Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.098007,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-819d-ad66-ebd03b6f65d4",
        "class": "G20",
        "model_evaluation": "14m-s2k-d256-c1000g250",
        "wws21_mean": 0.098007,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:09Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d256-c1000g250_GOOG_3864373/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d256-c1000g250_GOOG_3864373/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:09Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd819dad66ebd03b6f65d4",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 876,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065/45110",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d256-c1000g250_GOOG_3864373.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065 (step 45110)",
            "checkpoint_step": 45110,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d256-c1000g250_GOOG_3864373.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/14m-s2k-d256-c1000g250_GOOG_3864373\"         --n_cond_msgs 1000         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 45110         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.139466+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 1000,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d256-c1000g250_GOOG_3864373.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704065_hpxxr5mb_3704065\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/14m-s2k-d256-c1000g250_GOOG_3864373\"         --n_cond_msgs 1000         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 45110         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d256-c1000g250_GOOG_3864373.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_14m-s2k-d256-c1000g250_GOOG_3864373.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/14m-s2k-d256-c1000g250_GOOG_3864373"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d256-c1000g250_GOOG_3864373/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d256-c1000g250_GOOG_3864373/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/14m-s2k-d256-c1000g250_GOOG_3864373/data_real.tar",
            "selector": null,
            "source_value": 0.098007,
            "report_record_id": "canonical-scores:37712c45-68fd-819d-ad66-ebd03b6f65d4",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-819e-ad39-f07c9ee71426",
      "source_row_id": "37712c45-68fd-819e-ad39-f07c9ee71426",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd819ead39f07c9ee71426",
      "run": "phaseB-m3-14m-s500-d128-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.135212,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.135212",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.155336,
      "reported_ks_context": 0.099879,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:32Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd819ead39f07c9ee71426",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:32Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:32Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:32Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.135212,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-819e-ad39-f07c9ee71426",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s500-d128-b-avg-c250g250",
        "wws21_mean": 0.135212,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:32Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:32Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd819ead39f07c9ee71426",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 877,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947/log_rank0.txt",
            "evidence_line": 621,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703791_os296u12_3703791",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703791_os296u12_3703791\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703791_os296u12_3703791 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947/log_rank0.txt",
                "evidence_line": 623,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791",
            "availability_checked_at": "2026-09-12T06:13:15.366404+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703791_os296u12_3703791/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703791_os296u12_3703791\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d128-b-avg-c250g250_GOOG_3713947/data_real.tar",
            "selector": null,
            "source_value": 0.135212,
            "report_record_id": "canonical-scores:37712c45-68fd-819e-ad39-f07c9ee71426",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81a0-8703-d4d87751e975",
      "source_row_id": "37712c45-68fd-81a0-8703-d4d87751e975",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81a08703d4d87751e975",
      "run": "rmsfix-78m-s33480-c250g250-absmax8k",
      "stock": null,
      "model_size": "78M",
      "run_family": "rmsfix",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.093188,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.093188",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.133658,
      "reported_ks_context": 0.086428,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:08Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81a08703d4d87751e975",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:08Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:08Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:08Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.093188,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81a0-8703-d4d87751e975",
        "class": "G20",
        "model_evaluation": "rmsfix-78m-s33480-c250g250-absmax8k",
        "wws21_mean": 0.093188,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:08Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:08Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81a08703d4d87751e975",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 878,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4184212_yhmpnuk6_4184212/33480",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4184212_yhmpnuk6_4184212 (step 33480)",
            "checkpoint_step": 33480,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4184212_yhmpnuk6_4184212",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4184212_yhmpnuk6_4184212\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 3136         --batch_size 32         --checkpoint_step 33480         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136_absmax8000_as_s500.txt\" --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.179943+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4184212_yhmpnuk6_4184212",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4184212_yhmpnuk6_4184212"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j4184212_yhmpnuk6_4184212\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 3136         --batch_size 32         --checkpoint_step 33480         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136_absmax8000_as_s500.txt\" --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/rmsfix-78m-s33480-c250g250-absmax8k_GOOG_4240192/data_real.tar",
            "selector": null,
            "source_value": 0.093188,
            "report_record_id": "canonical-scores:37712c45-68fd-81a0-8703-d4d87751e975",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81a0-922c-d13acc56d4a6",
      "source_row_id": "37712c45-68fd-81a0-922c-d13acc56d4a6",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81a0922cd13acc56d4a6",
      "run": "H2-55M-ctx4k-cond4500",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.116645,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.116645",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.112012,
      "reported_ks_context": 0.074032,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:41Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81a0922cd13acc56d4a6",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:41Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:41Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:41Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.116645,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81a0-922c-d13acc56d4a6",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond4500",
        "wws21_mean": 0.116645,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:41Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond4500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond4500_shard7_integrated_2632447.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond4500_GOOG_2632447/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:41Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond4500_GOOG_2632447/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81a0922cd13acc56d4a6",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 879,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond4500_GOOG_2632447/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 4500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond4500_GOOG_2632447.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond4500_GOOG_2632447\"         --n_cond_msgs 4500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 6         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond4500_GOOG_2632447.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond4500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond4500_shard7_integrated_2632447.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond4500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond4500_integrated_2632447.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1166445169321702,
        "feature_points": {
          "spread": 0.1335781667147216,
          "ask_cancellation_depth": 0.23167531645742978,
          "bid_cancellation_depth": 0.31929074130827817,
          "limit_ask_order_levels": 0.019773777873856323,
          "limit_bid_order_levels": 0.027675373309739505,
          "ask_cancellation_levels": 0.041894495103822094,
          "bid_cancellation_levels": 0.042711453288408835,
          "vol_per_min": 0.14264638244688946,
          "ofi": 0.04060654142829484,
          "ofi_up": 0.06701071258679404,
          "ofi_stay": 0.04071302886707229,
          "orderbook_imbalance": 0.10121414336785335,
          "ofi_down": 0.05668961824710774,
          "log_inter_arrival_time": 0.03158279817429225,
          "log_time_to_cancel": 0.35146525796240513,
          "ask_volume_touch": 0.10137399726751752,
          "bid_volume_touch": 0.04241666576372634,
          "ask_volume": 0.0967076862284466,
          "bid_volume": 0.07935922606618881,
          "limit_ask_order_depth": 0.15893964190443938,
          "limit_bid_order_depth": 0.3222098312082903
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond4500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond4500_integrated_2632447.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.116645,
            "report_record_id": "canonical-scores:37712c45-68fd-81a0-922c-d13acc56d4a6",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81a1-846a-e6dfedde64f0",
      "source_row_id": "37712c45-68fd-81a1-846a-e6dfedde64f0",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81a1846ae6dfedde64f0",
      "run": "j2633975-goog-L100",
      "stock": null,
      "model_size": null,
      "run_family": "j2633975",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.273067,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.273067",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.171359,
      "reported_ks_context": 0.138175,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-05T12:20:22Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2118",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-05T12:20:22Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-05T12:20:22Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-05T12:20:22Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.273067,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81a1-846a-e6dfedde64f0",
        "class": "G20",
        "model_evaluation": "j2633975-goog-L100",
        "wws21_mean": 0.273067,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-05T12:20:22Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L100_GOOG_2930533/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L100_GOOG_2930533/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-05T12:20:22Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81a1846ae6dfedde64f0",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 880,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L100_GOOG_2930533.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975 (step 65073)",
            "checkpoint_step": 65073,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L100_GOOG_2930533.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5_l100eval         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/j2633975-goog-L100_GOOG_2930533\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 65073         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262799+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.121665+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L100_GOOG_2930533.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5_l100eval         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/j2633975-goog-L100_GOOG_2930533\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 65073         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L100_GOOG_2930533.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2633975-goog-L100_GOOG_2930533.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/j2633975-goog-L100_GOOG_2930533"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L100_GOOG_2930533/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L100_GOOG_2930533/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2633975-goog-L100_GOOG_2930533/data_real.tar",
            "selector": null,
            "source_value": 0.273067,
            "report_record_id": "canonical-scores:37712c45-68fd-81a1-846a-e6dfedde64f0",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81a3-9103-fef84fe917f9",
      "source_row_id": "37712c45-68fd-81a3-9103-fef84fe917f9",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81a39103fef84fe917f9",
      "run": "ticks-mamba2-soup",
      "stock": null,
      "model_size": null,
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.181342,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.181342",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.137973,
      "reported_ks_context": 0.094421,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:36Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81a39103fef84fe917f9",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:36Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:36Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:36Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.181342,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81a3-9103-fef84fe917f9",
        "class": "G20",
        "model_evaluation": "ticks-mamba2-soup",
        "wws21_mean": 0.181342,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:36Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-mamba2-soup/scores/scores_uncond_GOOG_ticks-mamba2-soup_integrated_3277518.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:36Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 3136 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81a39103fef84fe917f9",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 881,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469/log_rank0.txt",
            "evidence_line": 534,
            "evidence_text": "[Checkpoint] Loading msgpack checkpoint step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10/0",
              "availability": "exists",
              "entries": [
                "metadata",
                "state.msgpack"
              ],
              "root_entries": [
                "0",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10/0",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup_GOOG_3277518.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup_GOOG_3277518.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup_GOOG_3277518.out",
            "line": 132,
            "text": "  Loading: scores_uncond_GOOG_ticks-mamba2-soup_integrated_3277518.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last10\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup10_GOOG_2917469.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup_GOOG_3277518.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup_GOOG_3277518.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup10_GOOG_2917469"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-mamba2-soup_GOOG_3277518.out",
            "line": 132,
            "text": "  Loading: scores_uncond_GOOG_ticks-mamba2-soup_integrated_3277518.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-mamba2-soup/scores/scores_uncond_GOOG_ticks-mamba2-soup_integrated_3277518.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-mamba2-soup/scores/scores_uncond_GOOG_ticks-mamba2-soup_integrated_3277518.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1813415488622559,
        "feature_points": {
          "ask_cancellation_depth": 0.3105653962158392,
          "vol_per_min": 0.058822879774380825,
          "bid_volume": 0.09673908618765506,
          "ofi_up": 0.03854867005195023,
          "ofi_down": 0.03918590006148727,
          "spread": 0.18578859019078242,
          "ask_cancellation_ticks": 0.3098733137300282,
          "bid_cancellation_depth": 0.3484138466789832,
          "limit_bid_order_depth": 0.3279729056230072,
          "orderbook_imbalance": 0.07849510932154945,
          "limit_bid_order_ticks": 0.327995980653105,
          "bid_cancellation_ticks": 0.34789560809581865,
          "limit_ask_order_depth": 0.2933263512595122,
          "limit_ask_order_ticks": 0.2932937327311145,
          "ask_volume_touch": 0.05559187774161205,
          "log_time_to_cancel": 0.3244206420813768,
          "ask_volume": 0.10736393250729281,
          "ofi": 0.06252883279598859,
          "ofi_stay": 0.06357594413713756,
          "bid_volume_touch": 0.05254070751623455,
          "log_inter_arrival_time": 0.0852332187525183
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-mamba2-soup/scores/scores_uncond_GOOG_ticks-mamba2-soup_integrated_3277518.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.181342,
            "report_record_id": "canonical-scores:37712c45-68fd-81a3-9103-fef84fe917f9",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81a8-9088-c5d10f2267a0",
      "source_row_id": "37712c45-68fd-81a8-9088-c5d10f2267a0",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81a89088c5d10f2267a0",
      "run": "s2k-d256-s48740",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.108781,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.108781",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.146427,
      "reported_ks_context": 0.090347,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:32Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81a89088c5d10f2267a0",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:32Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:32Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:32Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.108781,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81a8-9088-c5d10f2267a0",
        "class": "G20",
        "model_evaluation": "s2k-d256-s48740",
        "wws21_mean": 0.108781,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:32Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s48740_GOOG_3835484/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s48740_GOOG_3835484/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:32Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81a89088c5d10f2267a0",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 882,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596/48740",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s48740_GOOG_3835484.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596 (step 48740)",
            "checkpoint_step": 48740,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s48740_GOOG_3835484.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s48740_GOOG_3835484\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 48740         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.121699+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s48740_GOOG_3835484.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-s48740_GOOG_3835484\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 48740         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s48740_GOOG_3835484.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-s48740_GOOG_3835484.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-d256-s48740_GOOG_3835484"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s48740_GOOG_3835484/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s48740_GOOG_3835484/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-s48740_GOOG_3835484/data_real.tar",
            "selector": null,
            "source_value": 0.108781,
            "report_record_id": "canonical-scores:37712c45-68fd-81a8-9088-c5d10f2267a0",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81a8-91ae-c38b3850045e",
      "source_row_id": "37712c45-68fd-81a8-91ae-c38b3850045e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81a891aec38b3850045e",
      "run": "R1-s4k-d256-raw-s42510",
      "stock": null,
      "model_size": null,
      "run_family": "R1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.173635,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.173635",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.175864,
      "reported_ks_context": 0.105157,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:53Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81a891aec38b3850045e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:53Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:53Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:53Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.173635,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81a8-91ae-c38b3850045e",
        "class": "G20",
        "model_evaluation": "R1-s4k-d256-raw-s42510",
        "wws21_mean": 0.173635,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:53Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s42510_GOOG_3745461/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s42510_GOOG_3745461/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:53Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81a891aec38b3850045e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 883,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096/42510",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s42510_GOOG_3745461.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096 (step 42510)",
            "checkpoint_step": 42510,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s42510_GOOG_3745461.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s42510_GOOG_3745461\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 42510         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.136864+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s42510_GOOG_3745461.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s42510_GOOG_3745461\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 42510         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s42510_GOOG_3745461.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s42510_GOOG_3745461.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s42510_GOOG_3745461"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s42510_GOOG_3745461/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s42510_GOOG_3745461/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s42510_GOOG_3745461/data_real.tar",
            "selector": null,
            "source_value": 0.173635,
            "report_record_id": "canonical-scores:37712c45-68fd-81a8-91ae-c38b3850045e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81a9-8771-cec0dcabc0d8",
      "source_row_id": "37712c45-68fd-81a9-8771-cec0dcabc0d8",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81a98771cec0dcabc0d8",
      "run": "mamba3-34m-eggroll-4n-P256-s300-v3",
      "stock": null,
      "model_size": "34M",
      "run_family": "mamba3",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.129802,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.129802",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.198091,
      "reported_ks_context": 0.105422,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:38Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81a98771cec0dcabc0d8",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:38Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:38Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:38Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.129802,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81a9-8771-cec0dcabc0d8",
        "class": "G20",
        "model_evaluation": "mamba3-34m-eggroll-4n-P256-s300-v3",
        "wws21_mean": 0.129802,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:38Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:38Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81a98771cec0dcabc0d8",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 884,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints (step 300)",
            "checkpoint_step": 300,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 300         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263181+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.181280+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 300         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
            "selector": null,
            "source_value": 0.129802,
            "report_record_id": "canonical-scores:37712c45-68fd-81a9-8771-cec0dcabc0d8",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81a9-937a-c382e790ae98",
      "source_row_id": "37712c45-68fd-81a9-937a-c382e790ae98",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81a9937ac382e790ae98",
      "run": "ctx4k-cond4000",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.151473,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.151473",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.125628,
      "reported_ks_context": 0.087641,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:44:00Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81a9937ac382e790ae98",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:44:00Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:44:00Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:44:00Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.151473,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81a9-937a-c382e790ae98",
        "class": "G20",
        "model_evaluation": "ctx4k-cond4000",
        "wws21_mean": 0.151473,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:44:00Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond4000/scores/scores_uncond_GOOG_ctx4k-cond4000_integrated_2828766.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond4000_GOOG_2828766/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:44:00Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond4000_GOOG_2828766/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81a9937ac382e790ae98",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 885,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond4000_GOOG_2828766/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 4000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond4000_GOOG_2828766.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond4000_GOOG_2828766\"         --n_cond_msgs 4000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond4000_GOOG_2828766.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond4000/scores/scores_uncond_GOOG_ctx4k-cond4000_integrated_2828766.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond4000/scores/scores_uncond_GOOG_ctx4k-cond4000_integrated_2828766.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15147346409453033,
        "feature_points": {
          "bid_cancellation_levels": 0.03220618512861555,
          "limit_ask_order_levels": 0.06833076432462608,
          "ofi_down": 0.06669442627907247,
          "spread": 0.28221424331859274,
          "vol_per_min": 0.04645350754853751,
          "ofi": 0.04496211017743292,
          "ofi_up": 0.047561776260209576,
          "ask_volume": 0.21908395503887845,
          "limit_bid_order_depth": 0.31707928251043216,
          "ask_cancellation_levels": 0.0587998463946132,
          "limit_bid_order_levels": 0.01999006679873367,
          "orderbook_imbalance": 0.11379702704419455,
          "limit_ask_order_depth": 0.1621889072161997,
          "ask_cancellation_depth": 0.3168011530252054,
          "bid_volume_touch": 0.11470983210899435,
          "bid_cancellation_depth": 0.33873523857740556,
          "ask_volume_touch": 0.11049549894290764,
          "bid_volume": 0.3250468360047897,
          "ofi_stay": 0.04636529922875544,
          "log_time_to_cancel": 0.39485891156110875,
          "log_inter_arrival_time": 0.05456787849583121
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond4000/scores/scores_uncond_GOOG_ctx4k-cond4000_integrated_2828766.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.151473,
            "report_record_id": "canonical-scores:37712c45-68fd-81a9-937a-c382e790ae98",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81a9-bd8b-efc0b5c8701f",
      "source_row_id": "37712c45-68fd-81a9-bd8b-efc0b5c8701f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81a9bd8befc0b5c8701f",
      "run": "j2504811_hfmn2548",
      "stock": null,
      "model_size": null,
      "run_family": "j2504811",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.231756,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.231756",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.189904,
      "reported_ks_context": 0.119551,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T22:00:39Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2174",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T22:00:39Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T22:00:39Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T22:00:39Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.231756,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81a9-bd8b-efc0b5c8701f",
        "class": "G20",
        "model_evaluation": "j2504811_hfmn2548",
        "wws21_mean": 0.231756,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T22:00:39Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_shard7_integrated_2566546.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546/data_real.tar;/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566544/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T22:00:39Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81a9bd8befc0b5c8701f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 886,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "192987",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546/log_rank0.txt",
            "evidence_line": 628,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "103505",
                "108205",
                "112915",
                "160000",
                "164710",
                "174140",
                "178853",
                "183557",
                "188270",
                "192987",
                "197705",
                "202424",
                "207134",
                "211843",
                "216558",
                "23310",
                "28025",
                "32740",
                "37455",
                "56320",
                "70465",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Multiple source input candidates were resolved by the exact score filename explicitly listed in the matching integrated-job log. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 520,
            "text": "  - /projects/s5e/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_shard7_integrated_2566546.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 192987         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 12,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 520,
            "text": "  - /projects/s5e/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_shard7_integrated_2566546.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_shard7_integrated_2566546.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_integrated_2566546.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.23175622688048572,
        "feature_points": {
          "spread": 0.4228574483487797,
          "ask_cancellation_depth": 0.450635592914468,
          "bid_cancellation_depth": 0.472347188503442,
          "limit_ask_order_levels": 0.1174546570881266,
          "limit_bid_order_levels": 0.10914834105981358,
          "ask_cancellation_levels": 0.16848615565439748,
          "bid_cancellation_levels": 0.13567124666286237,
          "vol_per_min": 0.13371054418951422,
          "ofi": 0.11802831948540084,
          "ofi_up": 0.15313352338481845,
          "ofi_stay": 0.12045877340386857,
          "orderbook_imbalance": 0.11082441992647245,
          "ofi_down": 0.10881454275305076,
          "log_inter_arrival_time": 0.20927945307560433,
          "log_time_to_cancel": 0.6670623874738916,
          "ask_volume_touch": 0.16926273641660505,
          "bid_volume_touch": 0.10257804957957822,
          "ask_volume": 0.11102227410808888,
          "bid_volume": 0.07750528565099786,
          "limit_ask_order_depth": 0.4521070202383495,
          "limit_bid_order_depth": 0.4564928045720696
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_integrated_2566546.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.231756,
            "report_record_id": "canonical-scores:37712c45-68fd-81a9-bd8b-efc0b5c8701f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81aa-84c1-d28293278fd9",
      "source_row_id": "37712c45-68fd-81aa-84c1-d28293278fd9",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81aa84c1d28293278fd9",
      "run": "adamw-32n-s16970",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.154837,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.154837",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.122011,
      "reported_ks_context": 0.077253,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:48Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81aa84c1d28293278fd9",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:48Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:48Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:48Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.154837,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81aa-84c1-d28293278fd9",
        "class": "G20",
        "model_evaluation": "adamw-32n-s16970",
        "wws21_mean": 0.154837,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:48Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s16970_GOOG_2871384/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s16970_GOOG_2871384/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:48Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81aa84c1d28293278fd9",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 887,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s16970_GOOG_2871384.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 16970)",
            "checkpoint_step": 16970,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s16970_GOOG_2871384.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s16970_GOOG_2871384\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 16970         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s16970_GOOG_2871384.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s16970_GOOG_2871384\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 16970         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s16970_GOOG_2871384.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s16970_GOOG_2871384.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s16970_GOOG_2871384"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s16970_GOOG_2871384/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s16970_GOOG_2871384/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s16970_GOOG_2871384/data_real.tar",
            "selector": null,
            "source_value": 0.154837,
            "report_record_id": "canonical-scores:37712c45-68fd-81aa-84c1-d28293278fd9",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81aa-ba05-fa7b7dcbe5f7",
      "source_row_id": "37712c45-68fd-81aa-ba05-fa7b7dcbe5f7",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81aaba05fa7b7dcbe5f7",
      "run": "H1-120M-cond3500",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.137108,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.137108",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.113063,
      "reported_ks_context": 0.078398,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:32:55Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81aaba05fa7b7dcbe5f7",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:32:55Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:32:55Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:32:55Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.137108,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81aa-ba05-fa7b7dcbe5f7",
        "class": "G20",
        "model_evaluation": "H1-120M-cond3500",
        "wws21_mean": 0.137108,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:32:55Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond3500/scores/scores_uncond_GOOG_H1-120M-cond3500_shard5_integrated_2598224.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond3500_GOOG_2598224/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:32:55Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond3500_GOOG_2598224/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81aaba05fa7b7dcbe5f7",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 888,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond3500_GOOG_2598224/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 3500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond3500_GOOG_2598224.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-120M-cond3500_GOOG_2598224\"         --n_cond_msgs 3500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 135458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond3500_GOOG_2598224.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond3500/scores/scores_uncond_GOOG_H1-120M-cond3500_shard5_integrated_2598224.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond3500/scores/scores_uncond_GOOG_H1-120M-cond3500_integrated_2598224.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1371077351746023,
        "feature_points": {
          "spread": 0.21748346061920573,
          "ask_cancellation_depth": 0.2864961882272763,
          "bid_cancellation_depth": 0.3137270383136615,
          "limit_ask_order_levels": 0.09426696730908735,
          "limit_bid_order_levels": 0.07668590796019283,
          "ask_cancellation_levels": 0.09450353321712544,
          "bid_cancellation_levels": 0.0740143954246124,
          "vol_per_min": 0.05412764957134378,
          "ofi": 0.06634836540908269,
          "ofi_up": 0.10636716048983595,
          "ofi_stay": 0.06475629101176453,
          "orderbook_imbalance": 0.09301882042719407,
          "ofi_down": 0.09384783125125203,
          "log_inter_arrival_time": 0.06341184305541957,
          "log_time_to_cancel": 0.3716087105933068,
          "ask_volume_touch": 0.06056857917913345,
          "bid_volume_touch": 0.03769026619102746,
          "ask_volume": 0.08362223386128277,
          "bid_volume": 0.049459320155014745,
          "limit_ask_order_depth": 0.24685368257570892,
          "limit_bid_order_depth": 0.3304041938241204
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond3500/scores/scores_uncond_GOOG_H1-120M-cond3500_integrated_2598224.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.137108,
            "report_record_id": "canonical-scores:37712c45-68fd-81aa-ba05-fa7b7dcbe5f7",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ab-84ad-d87775de2d5a",
      "source_row_id": "37712c45-68fd-81ab-84ad-d87775de2d5a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ab84add87775de2d5a",
      "run": "adamw-s71307",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.153968,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.153968",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.119958,
      "reported_ks_context": 0.083568,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:46Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81ab84add87775de2d5a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:46Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:46Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:46Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.153968,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ab-84ad-d87775de2d5a",
        "class": "G20",
        "model_evaluation": "adamw-s71307",
        "wws21_mean": 0.153968,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:46Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s71307_GOOG_2868487/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s71307_GOOG_2868487/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:46Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ab84add87775de2d5a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 889,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s71307_GOOG_2868487.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661 (step 71307)",
            "checkpoint_step": 71307,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s71307_GOOG_2868487.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s71307_GOOG_2868487\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 71307         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263172+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.179826+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s71307_GOOG_2868487.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s71307_GOOG_2868487\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 71307         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s71307_GOOG_2868487.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s71307_GOOG_2868487.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-s71307_GOOG_2868487"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s71307_GOOG_2868487/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s71307_GOOG_2868487/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s71307_GOOG_2868487/data_real.tar",
            "selector": null,
            "source_value": 0.153968,
            "report_record_id": "canonical-scores:37712c45-68fd-81ab-84ad-d87775de2d5a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ab-9313-fe588fa8a296",
      "source_row_id": "37712c45-68fd-81ab-9313-fe588fa8a296",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ab9313fe588fa8a296",
      "run": "H1-75M-38k",
      "stock": null,
      "model_size": "75M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.147853,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.147853",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.130101,
      "reported_ks_context": 0.078863,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:24Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81ab9313fe588fa8a296",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:24Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:24Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:24Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.147853,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ab-9313-fe588fa8a296",
        "class": "G20",
        "model_evaluation": "H1-75M-38k",
        "wws21_mean": 0.147853,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:24Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-75M-38k/scores/scores_uncond_GOOG_H1-75M-38k_integrated_2572577.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-75M-38k_GOOG_2572577/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:24Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-75M-38k_GOOG_2572577/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ab9313fe588fa8a296",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 890,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "38002",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099/38002",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-75M-38k_GOOG_2572577/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099/38002/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099/38002",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099/38002",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "29130",
                "33557",
                "35780",
                "38002",
                "40216",
                "42425",
                "44634",
                "46840",
                "49053",
                "51266",
                "53483",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099/38002",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-75M-38k_GOOG_2572577.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2496099_7jq7hw9s_2496099\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-75M-38k_GOOG_2572577\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 38002         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-75M-38k_GOOG_2572577.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-75M-38k/scores/scores_uncond_GOOG_H1-75M-38k_integrated_2572577.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-75M-38k/scores/scores_uncond_GOOG_H1-75M-38k_integrated_2572577.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14785338648480156,
        "feature_points": {
          "spread": 0.3553781733946619,
          "ask_cancellation_depth": 0.3108250143603727,
          "bid_cancellation_depth": 0.33096842084831385,
          "limit_ask_order_levels": 0.04446384775756405,
          "limit_bid_order_levels": 0.03245076515812615,
          "ask_cancellation_levels": 0.07912740304750976,
          "bid_cancellation_levels": 0.09104927306314901,
          "vol_per_min": 0.07274944054253829,
          "ofi": 0.10240456485513875,
          "ofi_up": 0.07644651966539728,
          "ofi_stay": 0.10332491679158509,
          "orderbook_imbalance": 0.1063711567142646,
          "ofi_down": 0.0664575686395907,
          "log_inter_arrival_time": 0.02369303116734683,
          "log_time_to_cancel": 0.28763674076491286,
          "ask_volume_touch": 0.06687427029728622,
          "bid_volume_touch": 0.06851255670687167,
          "ask_volume": 0.10147939195786324,
          "bid_volume": 0.12103174960466181,
          "limit_ask_order_depth": 0.3165034880199054,
          "limit_bid_order_depth": 0.34717282282377276
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-75M-38k/scores/scores_uncond_GOOG_H1-75M-38k_integrated_2572577.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.147853,
            "report_record_id": "canonical-scores:37712c45-68fd-81ab-9313-fe588fa8a296",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ad-8cf6-ca907469bfa3",
      "source_row_id": "37712c45-68fd-81ad-8cf6-ca907469bfa3",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ad8cf6ca907469bfa3",
      "run": "H2-55M-ctx4k-cond2500",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.129681,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.129681",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.111429,
      "reported_ks_context": 0.072294,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81ad8cf6ca907469bfa3",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.129681,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ad-8cf6-ca907469bfa3",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond2500",
        "wws21_mean": 0.129681,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond2500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond2500_shard8_integrated_2617327.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond2500_GOOG_2617327/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond2500_GOOG_2617327/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ad8cf6ca907469bfa3",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 891,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond2500_GOOG_2617327/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 2500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond2500_GOOG_2617327.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond2500_GOOG_2617327\"         --n_cond_msgs 2500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond2500_GOOG_2617327.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond2500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond2500_shard8_integrated_2617327.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond2500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond2500_integrated_2617327.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.12968063996993962,
        "feature_points": {
          "spread": 0.14384499231091077,
          "ask_cancellation_depth": 0.21271503431674152,
          "bid_cancellation_depth": 0.3040007468374106,
          "limit_ask_order_levels": 0.02724652472811574,
          "limit_bid_order_levels": 0.03589944788083181,
          "ask_cancellation_levels": 0.04057123163864545,
          "bid_cancellation_levels": 0.023454435545342493,
          "vol_per_min": 0.07316280578017301,
          "ofi": 0.05679740458558976,
          "ofi_up": 0.13441741873730248,
          "ofi_stay": 0.05674188085354271,
          "orderbook_imbalance": 0.10483402090686492,
          "ofi_down": 0.061743352763987305,
          "log_inter_arrival_time": 0.03321062660313135,
          "log_time_to_cancel": 0.35156936361882574,
          "ask_volume_touch": 0.08510676314031299,
          "bid_volume_touch": 0.11339149957616414,
          "ask_volume": 0.08983192546734269,
          "bid_volume": 0.13891859016340544,
          "limit_ask_order_depth": 0.31272469816208537,
          "limit_bid_order_depth": 0.32311067575200597
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond2500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond2500_integrated_2617327.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.129681,
            "report_record_id": "canonical-scores:37712c45-68fd-81ad-8cf6-ca907469bfa3",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ad-9d48-f02334f7cf4f",
      "source_row_id": "37712c45-68fd-81ad-9d48-f02334f7cf4f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ad9d48f02334f7cf4f",
      "run": "muon-s84432",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.240032,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.240032",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.17913,
      "reported_ks_context": 0.121025,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:58Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81ad9d48f02334f7cf4f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:58Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:58Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:58Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.240032,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ad-9d48-f02334f7cf4f",
        "class": "G20",
        "model_evaluation": "muon-s84432",
        "wws21_mean": 0.240032,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:58Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s84432_GOOG_2868875/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s84432_GOOG_2868875/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:58Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ad9d48f02334f7cf4f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 892,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s84432_GOOG_2868875.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248 (step 84432)",
            "checkpoint_step": 84432,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s84432_GOOG_2868875.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s84432_GOOG_2868875\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 84432         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263142+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.169342+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s84432_GOOG_2868875.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2815248_4tkbytct_2815248\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s84432_GOOG_2868875\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 84432         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s84432_GOOG_2868875.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s84432_GOOG_2868875.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/muon-s84432_GOOG_2868875"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s84432_GOOG_2868875/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s84432_GOOG_2868875/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s84432_GOOG_2868875/data_real.tar",
            "selector": null,
            "source_value": 0.240032,
            "report_record_id": "canonical-scores:37712c45-68fd-81ad-9d48-f02334f7cf4f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ae-a497-ed3d103a438a",
      "source_row_id": "37712c45-68fd-81ae-a497-ed3d103a438a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81aea497ed3d103a438a",
      "run": "logical-serenity-19_jan2026",
      "stock": null,
      "model_size": null,
      "run_family": "logical",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.386747,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.386747",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.267177,
      "reported_ks_context": 0.233797,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:12Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81aea497ed3d103a438a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:12Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:12Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:12Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.386747,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ae-a497-ed3d103a438a",
        "class": "G20",
        "model_evaluation": "logical-serenity-19_jan2026",
        "wws21_mean": 0.386747,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:12Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_logical-serenity-19_jan2026/scores/scores_uncond_GOOG_logical-serenity-19_jan2026_shard7_integrated_2516810.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/logical-serenity-19_jan2026_GOOG_2516810/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:12Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/logical-serenity-19_jan2026_GOOG_2516810/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81aea497ed3d103a438a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 893,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "89410",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/logical-serenity-19_4dhsl6me/89410",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/logical-serenity-19_4dhsl6me",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/logical-serenity-19_jan2026_GOOG_2516810/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/logical-serenity-19_4dhsl6me/89410/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/logical-serenity-19_4dhsl6me/89410",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/LOBS5/checkpoints/logical-serenity-19_4dhsl6me/89410",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "16295",
                "27230",
                "35170",
                "48610",
                "58500",
                "70191",
                "72234",
                "74276",
                "76312",
                "78358",
                "80402",
                "82435",
                "84470",
                "87461",
                "89410",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/LOBS5/checkpoints/logical-serenity-19_4dhsl6me/89410",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_logical-serenity-19_jan2026_GOOG_2516810.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/logical-serenity-19_4dhsl6me\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/logical-serenity-19_jan2026_GOOG_2516810\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 89410         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_logical-serenity-19_jan2026_GOOG_2516810.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_logical-serenity-19_jan2026/scores/scores_uncond_GOOG_logical-serenity-19_jan2026_shard7_integrated_2516810.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_logical-serenity-19_jan2026/scores/scores_uncond_GOOG_logical-serenity-19_jan2026_integrated_2516810.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.3867471633596748,
        "feature_points": {
          "spread": 0.6953986970161165,
          "ask_cancellation_depth": 0.6914663601781437,
          "bid_cancellation_depth": 0.672414985887453,
          "limit_ask_order_levels": 0.36595686432476243,
          "limit_bid_order_levels": 0.32985813501181394,
          "ask_cancellation_levels": 0.37520839643108367,
          "bid_cancellation_levels": 0.3582268362307641,
          "vol_per_min": 0.036218978059724724,
          "ofi": 0.20108904180161638,
          "ofi_up": 0.23832722329404424,
          "ofi_stay": 0.2020430846416981,
          "orderbook_imbalance": 0.13510585186789664,
          "ofi_down": 0.20648340689443806,
          "log_inter_arrival_time": 0.5753561091412516,
          "log_time_to_cancel": 0.15532598542504966,
          "ask_volume_touch": 0.1751629975658617,
          "bid_volume_touch": 0.1758257401301127,
          "ask_volume": 0.6275793938765205,
          "bid_volume": 0.5765812616401798,
          "limit_ask_order_depth": 0.6936529836984395,
          "limit_bid_order_depth": 0.6344080974362002
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_logical-serenity-19_jan2026/scores/scores_uncond_GOOG_logical-serenity-19_jan2026_integrated_2516810.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.386747,
            "report_record_id": "canonical-scores:37712c45-68fd-81ae-a497-ed3d103a438a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b0-b132-f12ca76feeea",
      "source_row_id": "37712c45-68fd-81b0-b132-f12ca76feeea",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b0b132f12ca76feeea",
      "run": "s2k-cooldown-s87760",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.085542,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.085542",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.148338,
      "reported_ks_context": 0.091143,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:07Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b0b132f12ca76feeea",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:07Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:07Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:07Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.085542,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b0-b132-f12ca76feeea",
        "class": "G20",
        "model_evaluation": "s2k-cooldown-s87760",
        "wws21_mean": 0.085542,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:07Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s87760_GOOG_3855693/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s87760_GOOG_3855693/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:07Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b0b132f12ca76feeea",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 894,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/87760",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s87760_GOOG_3855693.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 87760)",
            "checkpoint_step": 87760,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s87760_GOOG_3855693.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-cooldown-s87760_GOOG_3855693\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 87760         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s87760_GOOG_3855693.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-cooldown-s87760_GOOG_3855693\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 87760         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s87760_GOOG_3855693.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cooldown-s87760_GOOG_3855693.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-cooldown-s87760_GOOG_3855693"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s87760_GOOG_3855693/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s87760_GOOG_3855693/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cooldown-s87760_GOOG_3855693/data_real.tar",
            "selector": null,
            "source_value": 0.085542,
            "report_record_id": "canonical-scores:37712c45-68fd-81b0-b132-f12ca76feeea",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b0-bfdc-e566b4f495d8",
      "source_row_id": "37712c45-68fd-81b0-bfdc-e566b4f495d8",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b0bfdce566b4f495d8",
      "run": "j2504167_step100378_gen2000",
      "stock": null,
      "model_size": null,
      "run_family": "j2504167",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.190878,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.190878",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.152376,
      "reported_ks_context": 0.114912,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T21:58:46Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2180",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T21:58:46Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T21:58:46Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T21:58:46Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.190878,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b0-bfdc-e566b4f495d8",
        "class": "G20",
        "model_evaluation": "j2504167_step100378_gen2000",
        "wws21_mean": 0.190878,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T21:58:46Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167_step100378_gen2000/scores/scores_uncond_GOOG_j2504167_step100378_gen2000_shard20_integrated_2541167.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167_step100378_gen2000_GOOG_2541167/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T21:58:46Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167_step100378_gen2000_GOOG_2541167/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b0bfdce566b4f495d8",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 895,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167_step100378_gen2000_GOOG_2541167/log_rank0.txt",
            "evidence_line": 659,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 2000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504167_step100378_gen2000_GOOG_2541167.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504167_step100378_gen2000_GOOG_2541167\"         --n_cond_msgs 500         --n_gen_msgs 2000         --n_sequences 1024         --batch_size 64         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504167_step100378_gen2000_GOOG_2541167.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167_step100378_gen2000/scores/scores_uncond_GOOG_j2504167_step100378_gen2000_shard20_integrated_2541167.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167_step100378_gen2000/scores/scores_uncond_GOOG_j2504167_step100378_gen2000_integrated_2541167.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1908784640638829,
        "feature_points": {
          "spread": 0.45156167363820504,
          "ask_cancellation_depth": 0.29715545792592507,
          "bid_cancellation_depth": 0.3569559848514857,
          "limit_ask_order_levels": 0.042017856053175484,
          "limit_bid_order_levels": 0.04422298154899931,
          "ask_cancellation_levels": 0.09157787845971649,
          "bid_cancellation_levels": 0.07883805674151062,
          "vol_per_min": 0.08996575934561067,
          "ofi": 0.06987055989929217,
          "ofi_up": 0.11218873791436111,
          "ofi_stay": 0.06883172017144171,
          "orderbook_imbalance": 0.10507143142129068,
          "ofi_down": 0.11276607761934965,
          "log_inter_arrival_time": 0.03349614815092746,
          "log_time_to_cancel": 0.34095295464720915,
          "ask_volume_touch": 0.18325270732893958,
          "bid_volume_touch": 0.12283826659300064,
          "ask_volume": 0.45832644780682363,
          "bid_volume": 0.31054330369479555,
          "limit_ask_order_depth": 0.3113318790360742,
          "limit_bid_order_depth": 0.32668186249340686
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167_step100378_gen2000/scores/scores_uncond_GOOG_j2504167_step100378_gen2000_integrated_2541167.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.190878,
            "report_record_id": "canonical-scores:37712c45-68fd-81b0-bfdc-e566b4f495d8",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b1-928b-d2c2e5fc3043",
      "source_row_id": "37712c45-68fd-81b1-928b-d2c2e5fc3043",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b1928bd2c2e5fc3043",
      "run": "N1b-split-s62183-L100sim",
      "stock": null,
      "model_size": null,
      "run_family": "N1b",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.219791,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.219791",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.200483,
      "reported_ks_context": 0.131319,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:34Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b1928bd2c2e5fc3043",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:34Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:34Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:34Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.219791,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b1-928b-d2c2e5fc3043",
        "class": "G20",
        "model_evaluation": "N1b-split-s62183-L100sim",
        "wws21_mean": 0.219791,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:34Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim_GOOG_3234654/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim_GOOG_3234654/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:34Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b1928bd2c2e5fc3043",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 896,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim_GOOG_3234654.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129 (step 62183)",
            "checkpoint_step": 62183,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim_GOOG_3234654.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1b-split-s62183-L100sim_GOOG_3234654\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 62183         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263186+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.181399+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim_GOOG_3234654.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1b-split-s62183-L100sim_GOOG_3234654\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 62183         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim_GOOG_3234654.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim_GOOG_3234654.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1b-split-s62183-L100sim_GOOG_3234654"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim_GOOG_3234654/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim_GOOG_3234654/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim_GOOG_3234654/data_real.tar",
            "selector": null,
            "source_value": 0.219791,
            "report_record_id": "canonical-scores:37712c45-68fd-81b1-928b-d2c2e5fc3043",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b3-8f0e-ea98f7e5ca22",
      "source_row_id": "37712c45-68fd-81b3-8f0e-ea98f7e5ca22",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b38f0eea98f7e5ca22",
      "run": "s2k-cd-s96650",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.103456,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.103456",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.15162,
      "reported_ks_context": 0.091113,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:00Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b38f0eea98f7e5ca22",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:00Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:00Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:00Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.103456,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b3-8f0e-ea98f7e5ca22",
        "class": "G20",
        "model_evaluation": "s2k-cd-s96650",
        "wws21_mean": 0.103456,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:00Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s96650_GOOG_3864036/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s96650_GOOG_3864036/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:00Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b38f0eea98f7e5ca22",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 897,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/96650",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s96650_GOOG_3864036.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 96650)",
            "checkpoint_step": 96650,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s96650_GOOG_3864036.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s96650_GOOG_3864036\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 96650         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s96650_GOOG_3864036.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s96650_GOOG_3864036\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 96650         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s96650_GOOG_3864036.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s96650_GOOG_3864036.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s96650_GOOG_3864036"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s96650_GOOG_3864036/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s96650_GOOG_3864036/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s96650_GOOG_3864036/data_real.tar",
            "selector": null,
            "source_value": 0.103456,
            "report_record_id": "canonical-scores:37712c45-68fd-81b3-8f0e-ea98f7e5ca22",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b3-9e15-cc8c634bc5f6",
      "source_row_id": "37712c45-68fd-81b3-9e15-cc8c634bc5f6",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b39e15cc8c634bc5f6",
      "run": "pareto-360m-adamw-s14230",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.23018,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.23018",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.172265,
      "reported_ks_context": 0.122504,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:44:03Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b39e15cc8c634bc5f6",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:44:03Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:44:03Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:44:03Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.23018,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b3-9e15-cc8c634bc5f6",
        "class": "G20",
        "model_evaluation": "pareto-360m-adamw-s14230",
        "wws21_mean": 0.23018,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:44:03Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s14230_GOOG_3253401/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s14230_GOOG_3253401/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:44:03Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b39e15cc8c634bc5f6",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 898,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s14230_GOOG_3253401.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747 (step 14230)",
            "checkpoint_step": 14230,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s14230_GOOG_3253401.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s14230_GOOG_3253401\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 14230         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263082+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158732+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s14230_GOOG_3253401.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s14230_GOOG_3253401\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 14230         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s14230_GOOG_3253401.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s14230_GOOG_3253401.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s14230_GOOG_3253401"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s14230_GOOG_3253401/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s14230_GOOG_3253401/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s14230_GOOG_3253401/data_real.tar",
            "selector": null,
            "source_value": 0.23018,
            "report_record_id": "canonical-scores:37712c45-68fd-81b3-9e15-cc8c634bc5f6",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b3-a297-c619bff95452",
      "source_row_id": "37712c45-68fd-81b3-a297-c619bff95452",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b3a297c619bff95452",
      "run": "ticks-ctx2k-rand",
      "stock": null,
      "model_size": null,
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.205783,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.205783",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.162899,
      "reported_ks_context": 0.110081,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:15Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b3a297c619bff95452",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:15Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:15Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:15Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.205783,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b3-a297-c619bff95452",
        "class": "G20",
        "model_evaluation": "ticks-ctx2k-rand",
        "wws21_mean": 0.205783,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:15Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx2k-rand/scores/scores_uncond_GOOG_ticks-ctx2k-rand_integrated_3278229.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:15Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b3a297c619bff95452",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 899,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141/log_rank0.txt",
            "evidence_line": 634,
            "evidence_text": "[Checkpoint] Loading step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx2k-rand_GOOG_3278229.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx2k-rand_GOOG_3278229.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx2k-rand_GOOG_3278229.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-ctx2k-rand_integrated_3278229.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt2k-rand_GOOG_2888141.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt2k-rand_GOOG_2888141.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx2k-rand_GOOG_3278229.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx2k-rand_GOOG_3278229.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-ctxt2k-rand_GOOG_2888141"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx2k-rand_GOOG_3278229.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-ctx2k-rand_integrated_3278229.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx2k-rand/scores/scores_uncond_GOOG_ticks-ctx2k-rand_integrated_3278229.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx2k-rand/scores/scores_uncond_GOOG_ticks-ctx2k-rand_integrated_3278229.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.20578289059795382,
        "feature_points": {
          "log_inter_arrival_time": 0.024587131164189237,
          "ofi_down": 0.051305225048182156,
          "spread": 0.29874340267402705,
          "vol_per_min": 0.15568103728509664,
          "limit_bid_order_ticks": 0.3368334261598963,
          "ofi_up": 0.041074512604589206,
          "limit_ask_order_ticks": 0.31092419525721987,
          "limit_bid_order_depth": 0.33685452083023143,
          "ask_volume_touch": 0.07858611670124871,
          "bid_volume_touch": 0.0855468866507425,
          "bid_cancellation_depth": 0.3270665234297867,
          "limit_ask_order_depth": 0.31104618748816687,
          "ask_cancellation_depth": 0.298643247427297,
          "ofi": 0.03333157398866049,
          "bid_cancellation_ticks": 0.32594868963980095,
          "ask_cancellation_ticks": 0.2976592443520776,
          "log_time_to_cancel": 0.4203697047989754,
          "orderbook_imbalance": 0.1064361979756195,
          "bid_volume": 0.24348469199517267,
          "ofi_stay": 0.035062176478589537,
          "ask_volume": 0.20225601060746035
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx2k-rand/scores/scores_uncond_GOOG_ticks-ctx2k-rand_integrated_3278229.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.205783,
            "report_record_id": "canonical-scores:37712c45-68fd-81b3-a297-c619bff95452",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b3-bcc7-e13678b85947",
      "source_row_id": "37712c45-68fd-81b3-bcc7-e13678b85947",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b3bcc7e13678b85947",
      "run": "genlen-ctx500-c500-g1000",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.239992,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.239992",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.191886,
      "reported_ks_context": 0.134268,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:51Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b3bcc7e13678b85947",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:51Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:51Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:51Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.239992,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b3-bcc7-e13678b85947",
        "class": "G20",
        "model_evaluation": "genlen-ctx500-c500-g1000",
        "wws21_mean": 0.239992,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:51Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1000_GOOG_3405775/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1000_GOOG_3405775/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:51Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b3bcc7e13678b85947",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 900,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1000_GOOG_3405775.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1000_GOOG_3405775.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx500-c500-g1000_GOOG_3405775\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262945+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.125757+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 1000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1000_GOOG_3405775.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_55M_86k_93k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx500-c500-g1000_GOOG_3405775\"         --n_cond_msgs 500         --n_gen_msgs 1000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1000_GOOG_3405775.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx500-c500-g1000_GOOG_3405775.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx500-c500-g1000_GOOG_3405775"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1000_GOOG_3405775/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1000_GOOG_3405775/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx500-c500-g1000_GOOG_3405775/data_real.tar",
            "selector": null,
            "source_value": 0.239992,
            "report_record_id": "canonical-scores:37712c45-68fd-81b3-bcc7-e13678b85947",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b5-ad82-c8e390549f0f",
      "source_row_id": "37712c45-68fd-81b5-ad82-c8e390549f0f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b5ad82c8e390549f0f",
      "run": "H6_MoE_551M_step38349",
      "stock": null,
      "model_size": "551M",
      "run_family": "H6",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.190621,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.190621",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.148731,
      "reported_ks_context": 0.094779,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:21Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b5ad82c8e390549f0f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:21Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:21Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:21Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.190621,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b5-ad82-c8e390549f0f",
        "class": "G20",
        "model_evaluation": "H6_MoE_551M_step38349",
        "wws21_mean": 0.190621,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:21Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6_MoE_551M_step38349/scores/scores_uncond_GOOG_H6_MoE_551M_step38349_shard1_integrated_2559904.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H6_MoE_551M_step38349_GOOG_2559904/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:21Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H6_MoE_551M_step38349_GOOG_2559904/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b5ad82c8e390549f0f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 901,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "38349",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H6_MoE_551M_step38349_GOOG_2559904/log_rank0.txt",
            "evidence_line": 855,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "13595",
                "16900",
                "25150",
                "28450",
                "29275",
                "30922",
                "31749",
                "32573",
                "33397",
                "34224",
                "35048",
                "35875",
                "36701",
                "37524",
                "38349",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H6_MoE_551M_step38349_GOOG_2559904.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H6_MoE_551M_step38349_GOOG_2559904\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 38349         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H6_MoE_551M_step38349_GOOG_2559904.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6_MoE_551M_step38349/scores/scores_uncond_GOOG_H6_MoE_551M_step38349_shard1_integrated_2559904.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6_MoE_551M_step38349/scores/scores_uncond_GOOG_H6_MoE_551M_step38349_integrated_2559904.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19062057131915336,
        "feature_points": {
          "spread": 0.2262389882012455,
          "ask_cancellation_depth": 0.38595512547251665,
          "bid_cancellation_depth": 0.4152932919997894,
          "limit_ask_order_levels": 0.05551748186300162,
          "limit_bid_order_levels": 0.08010374620370145,
          "ask_cancellation_levels": 0.14500969218985965,
          "bid_cancellation_levels": 0.12392317231737429,
          "vol_per_min": 0.19107352827973256,
          "ofi": 0.11725877434145382,
          "ofi_up": 0.05404582756152286,
          "ofi_stay": 0.11970207963675288,
          "orderbook_imbalance": 0.09563487023614874,
          "ofi_down": 0.094466449067225,
          "log_inter_arrival_time": 0.13982039969547258,
          "log_time_to_cancel": 0.5747353620732184,
          "ask_volume_touch": 0.0913423729380413,
          "bid_volume_touch": 0.10304154981663279,
          "ask_volume": 0.06066820557342035,
          "bid_volume": 0.09997494949545536,
          "limit_ask_order_depth": 0.4106661715305014,
          "limit_bid_order_depth": 0.41855995920915356
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6_MoE_551M_step38349/scores/scores_uncond_GOOG_H6_MoE_551M_step38349_integrated_2559904.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.190621,
            "report_record_id": "canonical-scores:37712c45-68fd-81b5-ad82-c8e390549f0f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b6-8d44-f2d14a00219c",
      "source_row_id": "37712c45-68fd-81b6-8d44-f2d14a00219c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b68d44f2d14a00219c",
      "run": "adamw-s80311",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.171961,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.171961",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.136458,
      "reported_ks_context": 0.085858,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:19Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b68d44f2d14a00219c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:19Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:19Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:19Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.171961,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b6-8d44-f2d14a00219c",
        "class": "G20",
        "model_evaluation": "adamw-s80311",
        "wws21_mean": 0.171961,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:19Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s80311_GOOG_2868488/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s80311_GOOG_2868488/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:19Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b68d44f2d14a00219c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 902,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s80311_GOOG_2868488.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661 (step 80311)",
            "checkpoint_step": 80311,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s80311_GOOG_2868488.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s80311_GOOG_2868488\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 80311         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263172+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.179826+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s80311_GOOG_2868488.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2782661_upsz4ijw_2782661\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s80311_GOOG_2868488\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 80311         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s80311_GOOG_2868488.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s80311_GOOG_2868488.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-s80311_GOOG_2868488"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s80311_GOOG_2868488/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s80311_GOOG_2868488/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s80311_GOOG_2868488/data_real.tar",
            "selector": null,
            "source_value": 0.171961,
            "report_record_id": "canonical-scores:37712c45-68fd-81b6-8d44-f2d14a00219c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b8-9354-e52164923dd4",
      "source_row_id": "37712c45-68fd-81b8-9354-e52164923dd4",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b89354e52164923dd4",
      "run": "phaseB-m3-14m-h032-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.147895,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.147895",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.206213,
      "reported_ks_context": 0.115438,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:34Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b89354e52164923dd4",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:34Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:34Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:34Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.147895,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b8-9354-e52164923dd4",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-h032-b-avg-c250g250",
        "wws21_mean": 0.147895,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:34Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:34Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b89354e52164923dd4",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 903,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925/log_rank0.txt",
            "evidence_line": 604,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704414_ri9x2wsp_3704414",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704414_ri9x2wsp_3704414\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704414_ri9x2wsp_3704414 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925/log_rank0.txt",
                "evidence_line": 606,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414",
            "availability_checked_at": "2026-09-12T06:13:15.377226+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704414_ri9x2wsp_3704414/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704414_ri9x2wsp_3704414\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h032-b-avg-c250g250_GOOG_3713925/data_real.tar",
            "selector": null,
            "source_value": 0.147895,
            "report_record_id": "canonical-scores:37712c45-68fd-81b8-9354-e52164923dd4",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81b9-b099-c1ccd7cb66ab",
      "source_row_id": "37712c45-68fd-81b9-b099-c1ccd7cb66ab",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81b9b099c1ccd7cb66ab",
      "run": "phaseA-r100-ext-c500g500",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.109916,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.109916",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.170769,
      "reported_ks_context": 0.103073,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:22Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81b9b099c1ccd7cb66ab",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:22Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:22Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:22Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.109916,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81b9-b099-c1ccd7cb66ab",
        "class": "G20",
        "model_evaluation": "phaseA-r100-ext-c500g500",
        "wws21_mean": 0.109916,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:22Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r100-ext-c500g500/scores/scores_uncond_GOOG_phaseA-r100-ext-c500g500_integrated_3861645.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:22Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81b9b099c1ccd7cb66ab",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 904,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "28980",
            "path_kind": "directory",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
            "historical_root": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973/log_rank0.txt",
            "evidence_line": 614,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "11380",
                "13580",
                "15780",
                "17980",
                "20180",
                "22380",
                "24580",
                "2580",
                "26780",
                "28980",
                "380",
                "4780",
                "6980",
                "9180",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348/28980",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c500g500_GOOG_3861645.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c500g500_GOOG_3861645.out",
            "line": 17,
            "text": "  Using existing: /projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c500g500_GOOG_3861645.out",
            "line": 683,
            "text": "  Loading: scores_uncond_GOOG_phaseA-r100-ext-c500g500_integrated_3861645.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c500g500_GOOG_3701973.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696348_ntfmg06p_3696348\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 28980         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-a-s28980-c500g500_GOOG_3701973.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c500g500_GOOG_3861645.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c500g500_GOOG_3861645.out",
            "line": 17,
            "text": "  Using existing: /projects/s5e/lob_pipeline/inference_results/phaseA-r100-a-s28980-c500g500_GOOG_3701973"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r100-ext-c500g500_GOOG_3861645.out",
            "line": 683,
            "text": "  Loading: scores_uncond_GOOG_phaseA-r100-ext-c500g500_integrated_3861645.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r100-ext-c500g500/scores/scores_uncond_GOOG_phaseA-r100-ext-c500g500_integrated_3861645.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r100-ext-c500g500/scores/scores_uncond_GOOG_phaseA-r100-ext-c500g500_integrated_3861645.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.10991620963020295,
        "feature_points": {
          "bid_cancellation_depth": 0.18390713482415952,
          "ofi_up": 0.1234264814000394,
          "limit_bid_order_ticks": 0.07794441515393292,
          "vol_per_min": 0.08328512195962218,
          "spread": 0.2952677461595961,
          "ofi_down": 0.1345902802907352,
          "limit_bid_order_depth": 0.07791432115576512,
          "bid_cancellation_ticks": 0.1827844088080425,
          "ask_cancellation_ticks": 0.07249974972069524,
          "limit_ask_order_ticks": 0.04852937205776566,
          "limit_ask_order_depth": 0.04734097059717814,
          "ask_volume_touch": 0.06801805843391844,
          "bid_volume_touch": 0.06962388261761561,
          "ask_cancellation_depth": 0.07265575090466145,
          "bid_volume": 0.16519157039291496,
          "orderbook_imbalance": 0.10530692820354336,
          "ofi_stay": 0.06179970489142888,
          "ofi": 0.061699032858147256,
          "log_time_to_cancel": 0.2406831623676809,
          "ask_volume": 0.12389123723286044,
          "log_inter_arrival_time": 0.011881072203958563
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_phaseA-r100-ext-c500g500/scores/scores_uncond_GOOG_phaseA-r100-ext-c500g500_integrated_3861645.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.109916,
            "report_record_id": "canonical-scores:37712c45-68fd-81b9-b099-c1ccd7cb66ab",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81bb-917e-e3f50efbb96b",
      "source_row_id": "37712c45-68fd-81bb-917e-e3f50efbb96b",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81bb917ee3f50efbb96b",
      "run": "s2k-cd-s110790",
      "stock": null,
      "model_size": null,
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.100496,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.100496",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.140894,
      "reported_ks_context": 0.09409,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:24Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81bb917ee3f50efbb96b",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:24Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:24Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:24Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.100496,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81bb-917e-e3f50efbb96b",
        "class": "G20",
        "model_evaluation": "s2k-cd-s110790",
        "wws21_mean": 0.100496,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:24Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s110790_GOOG_3864040/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s110790_GOOG_3864040/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:24Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81bb917ee3f50efbb96b",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 905,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782/110790",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s110790_GOOG_3864040.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782 (step 110790)",
            "checkpoint_step": 110790,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s110790_GOOG_3864040.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s110790_GOOG_3864040\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 110790         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.125780+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s110790_GOOG_3864040.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3848782_zu75ac4u_3848782\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s110790_GOOG_3864040\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 110790         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s110790_GOOG_3864040.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-cd-s110790_GOOG_3864040.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/s2k-cd-s110790_GOOG_3864040"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s110790_GOOG_3864040/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s110790_GOOG_3864040/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-cd-s110790_GOOG_3864040/data_real.tar",
            "selector": null,
            "source_value": 0.100496,
            "report_record_id": "canonical-scores:37712c45-68fd-81bb-917e-e3f50efbb96b",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81bd-a9d7-edc575ce5adb",
      "source_row_id": "37712c45-68fd-81bd-a9d7-edc575ce5adb",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81bda9d7edc575ce5adb",
      "run": "pareto-360m-muon-s22184",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.206026,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.206026",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.159002,
      "reported_ks_context": 0.110035,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:12Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81bda9d7edc575ce5adb",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:12Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:12Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:12Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.206026,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81bd-a9d7-edc575ce5adb",
        "class": "G20",
        "model_evaluation": "pareto-360m-muon-s22184",
        "wws21_mean": 0.206026,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:12Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s22184_GOOG_3253412/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s22184_GOOG_3253412/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:12Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81bda9d7edc575ce5adb",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 906,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s22184_GOOG_3253412.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395 (step 22184)",
            "checkpoint_step": 22184,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s22184_GOOG_3253412.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s22184_GOOG_3253412\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 22184         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263190+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.181506+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s22184_GOOG_3253412.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s22184_GOOG_3253412\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 22184         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s22184_GOOG_3253412.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s22184_GOOG_3253412.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s22184_GOOG_3253412"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s22184_GOOG_3253412/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s22184_GOOG_3253412/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s22184_GOOG_3253412/data_real.tar",
            "selector": null,
            "source_value": 0.206026,
            "report_record_id": "canonical-scores:37712c45-68fd-81bd-a9d7-edc575ce5adb",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81bd-b630-ebcc931c733e",
      "source_row_id": "37712c45-68fd-81bd-b630-ebcc931c733e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81bdb630ebcc931c733e",
      "run": "H2-55M-ctx4k-cond1500",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.14237,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.14237",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.116625,
      "reported_ks_context": 0.078083,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:37Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81bdb630ebcc931c733e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:37Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:37Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:37Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.14237,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81bd-b630-ebcc931c733e",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond1500",
        "wws21_mean": 0.14237,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:37Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond1500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond1500_shard5_integrated_2617325.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond1500_GOOG_2617325/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:37Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond1500_GOOG_2617325/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81bdb630ebcc931c733e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 907,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond1500_GOOG_2617325/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond1500_GOOG_2617325.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond1500_GOOG_2617325\"         --n_cond_msgs 1500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond1500_GOOG_2617325.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond1500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond1500_shard5_integrated_2617325.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond1500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond1500_integrated_2617325.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14237028871276766,
        "feature_points": {
          "spread": 0.17372884165163638,
          "ask_cancellation_depth": 0.33544558627675203,
          "bid_cancellation_depth": 0.335133642647843,
          "limit_ask_order_levels": 0.024355524316856697,
          "limit_bid_order_levels": 0.02610152137659107,
          "ask_cancellation_levels": 0.03919835001061109,
          "bid_cancellation_levels": 0.050962220268608485,
          "vol_per_min": 0.1542053610082178,
          "ofi": 0.03871354977632288,
          "ofi_up": 0.08370930060478973,
          "ofi_stay": 0.03861722421819184,
          "orderbook_imbalance": 0.09910040575433579,
          "ofi_down": 0.09859716573871008,
          "log_inter_arrival_time": 0.03863244473952828,
          "log_time_to_cancel": 0.34362051056721576,
          "ask_volume_touch": 0.13530571119330043,
          "bid_volume_touch": 0.10549277460934972,
          "ask_volume": 0.09249613729054144,
          "bid_volume": 0.14262745723768216,
          "limit_ask_order_depth": 0.3034588236113679,
          "limit_bid_order_depth": 0.3302735100696683
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond1500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond1500_integrated_2617325.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.14237,
            "report_record_id": "canonical-scores:37712c45-68fd-81bd-b630-ebcc931c733e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c0-aa82-f6eb09f2c762",
      "source_row_id": "37712c45-68fd-81c0-aa82-f6eb09f2c762",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c0aa82f6eb09f2c762",
      "run": "j2505830_s5p7vl3x",
      "stock": null,
      "model_size": null,
      "run_family": "j2505830",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.232033,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.232033",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.170511,
      "reported_ks_context": 0.11277,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-28T06:57:57Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2163",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-28T06:57:57Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-28T06:57:57Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-28T06:57:57Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.232033,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c0-aa82-f6eb09f2c762",
        "class": "G20",
        "model_evaluation": "j2505830_s5p7vl3x",
        "wws21_mean": 0.232033,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-28T06:57:57Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_s5p7vl3x_GOOG_3151389/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_s5p7vl3x_GOOG_3151389/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-28T06:57:57Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c0aa82f6eb09f2c762",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 908,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_s5p7vl3x_GOOG_3151389.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830 (step 38349)",
            "checkpoint_step": 38349,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_s5p7vl3x_GOOG_3151389.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/j2505830_s5p7vl3x_GOOG_3151389\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 38349         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.166304+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_s5p7vl3x_GOOG_3151389.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/j2505830_s5p7vl3x_GOOG_3151389\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 38349         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_s5p7vl3x_GOOG_3151389.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_s5p7vl3x_GOOG_3151389.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/j2505830_s5p7vl3x_GOOG_3151389"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_s5p7vl3x_GOOG_3151389/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_s5p7vl3x_GOOG_3151389/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_s5p7vl3x_GOOG_3151389/data_real.tar",
            "selector": null,
            "source_value": 0.232033,
            "report_record_id": "canonical-scores:37712c45-68fd-81c0-aa82-f6eb09f2c762",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c3-81cf-d36c7f8fa778",
      "source_row_id": "37712c45-68fd-81c3-81cf-d36c7f8fa778",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c381cfd36c7f8fa778",
      "run": "phaseA-r000-a-s35690-c250g250",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.174764,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.174764",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.163857,
      "reported_ks_context": 0.105315,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:37Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c381cfd36c7f8fa778",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:37Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:37Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:37Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.174764,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c3-81cf-d36c7f8fa778",
        "class": "G20",
        "model_evaluation": "phaseA-r000-a-s35690-c250g250",
        "wws21_mean": 0.174764,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:37Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:37Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c381cfd36c7f8fa778",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 909,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c250g250_GOOG_3701966.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345 (step 35690)",
            "checkpoint_step": 35690,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c250g250_GOOG_3701966.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 35690         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.181744+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c250g250_GOOG_3701966.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 35690         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c250g250_GOOG_3701966.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c250g250_GOOG_3701966.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c250g250_GOOG_3701966/data_real.tar",
            "selector": null,
            "source_value": 0.174764,
            "report_record_id": "canonical-scores:37712c45-68fd-81c3-81cf-d36c7f8fa778",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c3-922f-feddd1e0df03",
      "source_row_id": "37712c45-68fd-81c3-922f-feddd1e0df03",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c3922ffeddd1e0df03",
      "run": "phaseB-m3-14m-s4000-d128-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.195243,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.195243",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.200401,
      "reported_ks_context": 0.110205,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:39Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c3922ffeddd1e0df03",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:39Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:39Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:39Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.195243,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c3-922f-feddd1e0df03",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s4000-d128-b-avg-c250g250",
        "wws21_mean": 0.195243,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:39Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:39Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c3922ffeddd1e0df03",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 910,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941/log_rank0.txt",
            "evidence_line": 608,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703800_w9cmkmrf_3703800",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703800_w9cmkmrf_3703800\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703800_w9cmkmrf_3703800 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941/log_rank0.txt",
                "evidence_line": 610,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800",
            "availability_checked_at": "2026-09-12T06:13:15.371164+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703800_w9cmkmrf_3703800\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c250g250_GOOG_3713941/data_real.tar",
            "selector": null,
            "source_value": 0.195243,
            "report_record_id": "canonical-scores:37712c45-68fd-81c3-922f-feddd1e0df03",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c3-a654-c0554cdf47e7",
      "source_row_id": "37712c45-68fd-81c3-a654-c0554cdf47e7",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c3a654c0554cdf47e7",
      "run": "soup-adamw-60k-84k",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.152886,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.152886",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.121333,
      "reported_ks_context": 0.079865,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:21Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c3a654c0554cdf47e7",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:21Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:21Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:21Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.152886,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c3-a654-c0554cdf47e7",
        "class": "G20",
        "model_evaluation": "soup-adamw-60k-84k",
        "wws21_mean": 0.152886,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:21Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_soup-adamw-60k-84k/scores/scores_uncond_GOOG_soup-adamw-60k-84k_integrated_2869520.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-adamw-60k-84k_GOOG_2869520/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:21Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-adamw-60k-84k_GOOG_2869520/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c3a654c0554cdf47e7",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 911,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_60k_84k",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_60k_84k",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-adamw-60k-84k_GOOG_2869520/log_rank0.txt",
            "evidence_line": 634,
            "evidence_text": "[Checkpoint] Loading step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_60k_84k (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_60k_84k",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_60k_84k/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_60k_84k",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_60k_84k/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-adamw-60k-84k_GOOG_2869520.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_adamw_60k_84k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-adamw-60k-84k_GOOG_2869520\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-adamw-60k-84k_GOOG_2869520.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_soup-adamw-60k-84k/scores/scores_uncond_GOOG_soup-adamw-60k-84k_integrated_2869520.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_soup-adamw-60k-84k/scores/scores_uncond_GOOG_soup-adamw-60k-84k_integrated_2869520.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1528856684145057,
        "feature_points": {
          "ask_cancellation_levels": 0.08240581650190865,
          "ask_cancellation_depth": 0.3314131536591333,
          "ask_volume": 0.13721257030456782,
          "limit_bid_order_levels": 0.09843835844528842,
          "limit_bid_order_depth": 0.36484196018128706,
          "bid_cancellation_levels": 0.0889659320310708,
          "limit_ask_order_depth": 0.34447790851739557,
          "limit_ask_order_levels": 0.08683568726160945,
          "vol_per_min": 0.055863595969939076,
          "ask_volume_touch": 0.05847732498561062,
          "ofi_up": 0.046239581091573834,
          "spread": 0.1852689743140286,
          "bid_cancellation_depth": 0.3540471790108386,
          "ofi_down": 0.07750336123429906,
          "orderbook_imbalance": 0.10569680022572618,
          "bid_volume": 0.13385662221082958,
          "ofi_stay": 0.10330659525004171,
          "ofi": 0.10229342300737251,
          "log_time_to_cancel": 0.3343702762851229,
          "bid_volume_touch": 0.04810639410082937,
          "log_inter_arrival_time": 0.07097752211614694
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_soup-adamw-60k-84k/scores/scores_uncond_GOOG_soup-adamw-60k-84k_integrated_2869520.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.152886,
            "report_record_id": "canonical-scores:37712c45-68fd-81c3-a654-c0554cdf47e7",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c3-b2ef-e31810f0a175",
      "source_row_id": "37712c45-68fd-81c3-b2ef-e31810f0a175",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c3b2efe31810f0a175",
      "run": "ctx2k-cond1000",
      "stock": null,
      "model_size": null,
      "run_family": "ctx2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.147981,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.147981",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.12088,
      "reported_ks_context": 0.083397,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:12Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c3b2efe31810f0a175",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:12Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:12Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:12Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.147981,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c3-b2ef-e31810f0a175",
        "class": "G20",
        "model_evaluation": "ctx2k-cond1000",
        "wws21_mean": 0.147981,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:12Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond1000/scores/scores_uncond_GOOG_ctx2k-cond1000_integrated_2826367.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond1000_GOOG_2826367/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:12Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond1000_GOOG_2826367/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c3b2efe31810f0a175",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 912,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "116460",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond1000_GOOG_2826367/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=116460 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950/116460",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "101180",
                "103865",
                "106554",
                "109240",
                "111923",
                "114608",
                "116460",
                "60915",
                "87740",
                "90425",
                "93115",
                "95804",
                "98489",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950/116460",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx2k-cond1000_GOOG_2826367.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx2k-cond1000_GOOG_2826367\"         --n_cond_msgs 1000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 116460         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx2k-cond1000_GOOG_2826367.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond1000/scores/scores_uncond_GOOG_ctx2k-cond1000_integrated_2826367.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond1000/scores/scores_uncond_GOOG_ctx2k-cond1000_integrated_2826367.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14798059786339457,
        "feature_points": {
          "bid_cancellation_depth": 0.30997882553204786,
          "bid_cancellation_levels": 0.0302002387619867,
          "ask_cancellation_depth": 0.25568616361190255,
          "limit_bid_order_depth": 0.2596671525162656,
          "ask_cancellation_levels": 0.029019292962342725,
          "ofi_stay": 0.044100662402642106,
          "ask_volume_touch": 0.10362497233752585,
          "vol_per_min": 0.171514393451313,
          "limit_ask_order_levels": 0.01606725801065051,
          "ofi": 0.04346819917209235,
          "limit_bid_order_levels": 0.02659935982940684,
          "ofi_up": 0.06552146782383227,
          "ofi_down": 0.07609495275095533,
          "spread": 0.2943275861340084,
          "limit_ask_order_depth": 0.30081331403706374,
          "ask_volume": 0.22922076441804234,
          "bid_volume_touch": 0.05534762366407263,
          "orderbook_imbalance": 0.10764764874270624,
          "bid_volume": 0.2726201578903797,
          "log_time_to_cancel": 0.3881927562061692,
          "log_inter_arrival_time": 0.02787976487587989
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond1000/scores/scores_uncond_GOOG_ctx2k-cond1000_integrated_2826367.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.147981,
            "report_record_id": "canonical-scores:37712c45-68fd-81c3-b2ef-e31810f0a175",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c3-b45d-ccbafe826bd2",
      "source_row_id": "37712c45-68fd-81c3-b45d-ccbafe826bd2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c3b45dccbafe826bd2",
      "run": "ctx2k-cond1500",
      "stock": null,
      "model_size": null,
      "run_family": "ctx2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.152392,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.152392",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.122879,
      "reported_ks_context": 0.086545,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:17Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c3b45dccbafe826bd2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:17Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:17Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:17Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.152392,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c3-b45d-ccbafe826bd2",
        "class": "G20",
        "model_evaluation": "ctx2k-cond1500",
        "wws21_mean": 0.152392,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:17Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond1500/scores/scores_uncond_GOOG_ctx2k-cond1500_integrated_2826368.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond1500_GOOG_2826368/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:17Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond1500_GOOG_2826368/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c3b45dccbafe826bd2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 913,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "116460",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond1500_GOOG_2826368/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=116460 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950/116460",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "101180",
                "103865",
                "106554",
                "109240",
                "111923",
                "114608",
                "116460",
                "60915",
                "87740",
                "90425",
                "93115",
                "95804",
                "98489",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950/116460",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx2k-cond1500_GOOG_2826368.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx2k-cond1500_GOOG_2826368\"         --n_cond_msgs 1500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 116460         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx2k-cond1500_GOOG_2826368.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond1500/scores/scores_uncond_GOOG_ctx2k-cond1500_integrated_2826368.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond1500/scores/scores_uncond_GOOG_ctx2k-cond1500_integrated_2826368.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15239191564469173,
        "feature_points": {
          "ask_cancellation_levels": 0.03250483790885468,
          "ofi_stay": 0.04404033608247365,
          "ofi_down": 0.09442897995282627,
          "limit_ask_order_depth": 0.2731777092524428,
          "bid_cancellation_levels": 0.03835560897023582,
          "vol_per_min": 0.09521403249203922,
          "ask_volume_touch": 0.09409187769261315,
          "ask_cancellation_depth": 0.3171507459742733,
          "bid_volume_touch": 0.06539115323225042,
          "bid_volume": 0.3228986289591019,
          "limit_ask_order_levels": 0.021803239032872793,
          "bid_cancellation_depth": 0.32431933945972063,
          "limit_bid_order_levels": 0.019603753152183073,
          "ofi_up": 0.08374997208109071,
          "ask_volume": 0.24879371420771468,
          "spread": 0.2723229722077988,
          "limit_bid_order_depth": 0.31052551631388137,
          "ofi": 0.04511042349347911,
          "log_time_to_cancel": 0.35879587112123235,
          "orderbook_imbalance": 0.1042016849140908,
          "log_inter_arrival_time": 0.03374983203735081
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond1500/scores/scores_uncond_GOOG_ctx2k-cond1500_integrated_2826368.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.152392,
            "report_record_id": "canonical-scores:37712c45-68fd-81c3-b45d-ccbafe826bd2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c3-b691-dbc89ee61131",
      "source_row_id": "37712c45-68fd-81c3-b691-dbc89ee61131",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c3b691dbc89ee61131",
      "run": "pareto-120m-s4072",
      "stock": null,
      "model_size": "120M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.268869,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.268869",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.197022,
      "reported_ks_context": 0.148806,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:06Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c3b691dbc89ee61131",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:06Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:06Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:06Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.268869,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c3-b691-dbc89ee61131",
        "class": "G20",
        "model_evaluation": "pareto-120m-s4072",
        "wws21_mean": 0.268869,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:06Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s4072_GOOG_3253393/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s4072_GOOG_3253393/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:06Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c3b691dbc89ee61131",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 914,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260/4072",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s4072_GOOG_3253393.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260 (step 4072)",
            "checkpoint_step": 4072,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s4072_GOOG_3253393.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s4072_GOOG_3253393\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 4072         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.145456+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s4072_GOOG_3253393.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2491260_lodxmy9q_2491260\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s4072_GOOG_3253393\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 4072         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s4072_GOOG_3253393.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s4072_GOOG_3253393.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s4072_GOOG_3253393"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s4072_GOOG_3253393/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s4072_GOOG_3253393/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s4072_GOOG_3253393/data_real.tar",
            "selector": null,
            "source_value": 0.268869,
            "report_record_id": "canonical-scores:37712c45-68fd-81c3-b691-dbc89ee61131",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c4-8d82-f6b180bbecbe",
      "source_row_id": "37712c45-68fd-81c4-8d82-f6b180bbecbe",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c48d82f6b180bbecbe",
      "run": "ctx2k-cond2000",
      "stock": null,
      "model_size": null,
      "run_family": "ctx2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.148668,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.148668",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.124092,
      "reported_ks_context": 0.084531,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:04Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c48d82f6b180bbecbe",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:04Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:04Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:04Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.148668,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c4-8d82-f6b180bbecbe",
        "class": "G20",
        "model_evaluation": "ctx2k-cond2000",
        "wws21_mean": 0.148668,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:04Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond2000/scores/scores_uncond_GOOG_ctx2k-cond2000_integrated_2826370.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond2000_GOOG_2826370/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:04Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond2000_GOOG_2826370/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c48d82f6b180bbecbe",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 915,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "116460",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx2k-cond2000_GOOG_2826370/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=116460 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950/116460",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "101180",
                "103865",
                "106554",
                "109240",
                "111923",
                "114608",
                "116460",
                "60915",
                "87740",
                "90425",
                "93115",
                "95804",
                "98489",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950/116460",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 2000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx2k-cond2000_GOOG_2826370.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale-2k/checkpoints/j2788950_7b4ikxua_2788950\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx2k-cond2000_GOOG_2826370\"         --n_cond_msgs 2000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 116460         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx2k-cond2000_GOOG_2826370.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond2000/scores/scores_uncond_GOOG_ctx2k-cond2000_integrated_2826370.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond2000/scores/scores_uncond_GOOG_ctx2k-cond2000_integrated_2826370.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14866811903175559,
        "feature_points": {
          "limit_ask_order_levels": 0.019735134550328517,
          "limit_bid_order_levels": 0.028391687993568296,
          "ask_cancellation_depth": 0.32668122524066545,
          "ofi_down": 0.08076363027964231,
          "bid_cancellation_depth": 0.32051299535624295,
          "limit_ask_order_depth": 0.26977523597308223,
          "ofi_up": 0.06484924557283715,
          "ask_volume_touch": 0.07703943474879653,
          "ask_cancellation_levels": 0.031130862504995943,
          "limit_bid_order_depth": 0.3062860473365168,
          "spread": 0.30899397371064896,
          "bid_cancellation_levels": 0.037598321768513084,
          "ofi": 0.03165652849406232,
          "vol_per_min": 0.1360763572286656,
          "log_time_to_cancel": 0.3758818370578297,
          "bid_volume": 0.21686575080934395,
          "bid_volume_touch": 0.061479919635047134,
          "ofi_stay": 0.032110227627214175,
          "ask_volume": 0.269900709478838,
          "orderbook_imbalance": 0.09309706680678251,
          "log_inter_arrival_time": 0.03320430749324575
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx2k-cond2000/scores/scores_uncond_GOOG_ctx2k-cond2000_integrated_2826370.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.148668,
            "report_record_id": "canonical-scores:37712c45-68fd-81c4-8d82-f6b180bbecbe",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c5-b073-f07af4cc61b7",
      "source_row_id": "37712c45-68fd-81c5-b073-f07af4cc61b7",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c5b073f07af4cc61b7",
      "run": "H1-120M-cond1000",
      "stock": null,
      "model_size": "120M",
      "run_family": "H1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.141841,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.141841",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.11397,
      "reported_ks_context": 0.079527,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:56Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c5b073f07af4cc61b7",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:56Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:56Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:56Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.141841,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c5-b073-f07af4cc61b7",
        "class": "G20",
        "model_evaluation": "H1-120M-cond1000",
        "wws21_mean": 0.141841,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:56Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1000/scores/scores_uncond_GOOG_H1-120M-cond1000_shard3_integrated_2598228.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond1000_GOOG_2598228/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:56Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond1000_GOOG_2598228/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c5b073f07af4cc61b7",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 916,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H1-120M-cond1000_GOOG_2598228/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 1000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond1000_GOOG_2598228.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H1-120M-cond1000_GOOG_2598228\"         --n_cond_msgs 1000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 8         --checkpoint_step 135458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H1-120M-cond1000_GOOG_2598228.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1000/scores/scores_uncond_GOOG_H1-120M-cond1000_shard3_integrated_2598228.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1000/scores/scores_uncond_GOOG_H1-120M-cond1000_integrated_2598228.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14184082238768164,
        "feature_points": {
          "spread": 0.18088947811128275,
          "ask_cancellation_depth": 0.2608926278526545,
          "bid_cancellation_depth": 0.31019668367599845,
          "limit_ask_order_levels": 0.08443493816859805,
          "limit_bid_order_levels": 0.055871890879748046,
          "ask_cancellation_levels": 0.08207240936314827,
          "bid_cancellation_levels": 0.0729340370243146,
          "vol_per_min": 0.15562012914367684,
          "ofi": 0.0827695713767402,
          "ofi_up": 0.07202674965323352,
          "ofi_stay": 0.08312299381316943,
          "orderbook_imbalance": 0.09739675291326395,
          "ofi_down": 0.10582613968226859,
          "log_inter_arrival_time": 0.06343111722561409,
          "log_time_to_cancel": 0.3756917451953855,
          "ask_volume_touch": 0.08808650635889878,
          "bid_volume_touch": 0.0688916423978562,
          "ask_volume": 0.07310447831660756,
          "bid_volume": 0.060411900381594746,
          "limit_ask_order_depth": 0.32638459216289095,
          "limit_bid_order_depth": 0.2786008864443694
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H1-120M-cond1000/scores/scores_uncond_GOOG_H1-120M-cond1000_integrated_2598228.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.141841,
            "report_record_id": "canonical-scores:37712c45-68fd-81c5-b073-f07af4cc61b7",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c7-867b-e3e3aa2fee30",
      "source_row_id": "37712c45-68fd-81c7-867b-e3e3aa2fee30",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c7867be3e3aa2fee30",
      "run": "phaseA-r000-a-s35690-c500g500",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.220829,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.220829",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.182706,
      "reported_ks_context": 0.123765,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:03Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c7867be3e3aa2fee30",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:03Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:03Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:03Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.220829,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c7-867b-e3e3aa2fee30",
        "class": "G20",
        "model_evaluation": "phaseA-r000-a-s35690-c500g500",
        "wws21_mean": 0.220829,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:03Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:03Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c7867be3e3aa2fee30",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 917,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345/35690",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c500g500_GOOG_3701967.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345 (step 35690)",
            "checkpoint_step": 35690,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c500g500_GOOG_3701967.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 35690         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.181744+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c500g500_GOOG_3701967.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696345_9xedylbq_3696345\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 35690         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c500g500_GOOG_3701967.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r000-a-s35690-c500g500_GOOG_3701967.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r000-a-s35690-c500g500_GOOG_3701967/data_real.tar",
            "selector": null,
            "source_value": 0.220829,
            "report_record_id": "canonical-scores:37712c45-68fd-81c7-867b-e3e3aa2fee30",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c8-82a8-f77d60ea4588",
      "source_row_id": "37712c45-68fd-81c8-82a8-f77d60ea4588",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c882a8f77d60ea4588",
      "run": "pareto-360m-muon-s28931",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.226632,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.226632",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.171145,
      "reported_ks_context": 0.121911,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:05Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c882a8f77d60ea4588",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:05Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:05Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:05Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.226632,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c8-82a8-f77d60ea4588",
        "class": "G20",
        "model_evaluation": "pareto-360m-muon-s28931",
        "wws21_mean": 0.226632,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:05Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s28931_GOOG_3253413/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s28931_GOOG_3253413/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:05Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c882a8f77d60ea4588",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 918,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s28931_GOOG_3253413.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395 (step 28931)",
            "checkpoint_step": 28931,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s28931_GOOG_3253413.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s28931_GOOG_3253413\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 28931         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263190+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.181506+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s28931_GOOG_3253413.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2889395_bogyzzls_2889395\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s28931_GOOG_3253413\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 28931         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s28931_GOOG_3253413.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s28931_GOOG_3253413.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s28931_GOOG_3253413"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s28931_GOOG_3253413/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s28931_GOOG_3253413/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s28931_GOOG_3253413/data_real.tar",
            "selector": null,
            "source_value": 0.226632,
            "report_record_id": "canonical-scores:37712c45-68fd-81c8-82a8-f77d60ea4588",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81c8-a158-c7f873b91a9c",
      "source_row_id": "37712c45-68fd-81c8-a158-c7f873b91a9c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81c8a158c7f873b91a9c",
      "run": "ticks-moe-pfr",
      "stock": null,
      "model_size": null,
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.228989,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.228989",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.189667,
      "reported_ks_context": 0.114169,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:35Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81c8a158c7f873b91a9c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:35Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:35Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:35Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.228989,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81c8-a158-c7f873b91a9c",
        "class": "G20",
        "model_evaluation": "ticks-moe-pfr",
        "wws21_mean": 0.228989,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:35Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-moe-pfr/scores/scores_uncond_GOOG_ticks-moe-pfr_integrated_3278226.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:35Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 3136 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81c8a158c7f873b91a9c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 919,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "38349",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214/log_rank0.txt",
            "evidence_line": 836,
            "evidence_text": "[Checkpoint] Loading step=38349 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "13595",
                "16900",
                "25150",
                "28450",
                "29275",
                "30922",
                "31749",
                "32573",
                "33397",
                "34224",
                "35048",
                "35875",
                "36701",
                "37524",
                "38349",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-moe-pfr_GOOG_3278226.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-moe-pfr_GOOG_3278226.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-moe-pfr_GOOG_3278226.out",
            "line": 132,
            "text": "  Loading: scores_uncond_GOOG_ticks-moe-pfr_integrated_3278226.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_pfr_GOOG_3152214.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 38349         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2505830_pfr_GOOG_3152214.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-moe-pfr_GOOG_3278226.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-moe-pfr_GOOG_3278226.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/j2505830_pfr_GOOG_3152214"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-moe-pfr_GOOG_3278226.out",
            "line": 132,
            "text": "  Loading: scores_uncond_GOOG_ticks-moe-pfr_integrated_3278226.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-moe-pfr/scores/scores_uncond_GOOG_ticks-moe-pfr_integrated_3278226.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-moe-pfr/scores/scores_uncond_GOOG_ticks-moe-pfr_integrated_3278226.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.22898921515256004,
        "feature_points": {
          "ofi_down": 0.03608781241305466,
          "ofi_up": 0.04381804978588523,
          "limit_bid_order_depth": 0.35519528986240306,
          "spread": 0.2615050304893132,
          "vol_per_min": 0.07119289363910616,
          "limit_ask_order_ticks": 0.3372098774902861,
          "ask_cancellation_depth": 0.3474197523669424,
          "ask_volume_touch": 0.12930722120190316,
          "limit_ask_order_depth": 0.33237163298027744,
          "bid_cancellation_depth": 0.38600214441317116,
          "limit_bid_order_ticks": 0.35517634278913623,
          "bid_cancellation_ticks": 0.38498238698252135,
          "ask_cancellation_ticks": 0.34715513001195925,
          "bid_volume_touch": 0.08060770589541032,
          "log_inter_arrival_time": 0.1118659445449865,
          "orderbook_imbalance": 0.09910679556058011,
          "log_time_to_cancel": 0.5442262849283119,
          "ofi": 0.08727479392442947,
          "ofi_stay": 0.09004068871911072,
          "ask_volume": 0.22068182038640896,
          "bid_volume": 0.1875459198185636
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-moe-pfr/scores/scores_uncond_GOOG_ticks-moe-pfr_integrated_3278226.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.228989,
            "report_record_id": "canonical-scores:37712c45-68fd-81c8-a158-c7f873b91a9c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ca-99cb-f8f57c371a21",
      "source_row_id": "37712c45-68fd-81ca-99cb-f8f57c371a21",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ca99cbf8f57c371a21",
      "run": "N1-base-s63328-L100",
      "stock": null,
      "model_size": null,
      "run_family": "N1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.170647,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.170647",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.159111,
      "reported_ks_context": 0.115775,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:01Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81ca99cbf8f57c371a21",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:01Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:01Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:01Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.170647,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ca-99cb-f8f57c371a21",
        "class": "G20",
        "model_evaluation": "N1-base-s63328-L100",
        "wws21_mean": 0.170647,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:01Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100_GOOG_3231570/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100_GOOG_3231570/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:01Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ca99cbf8f57c371a21",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 920,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391/63328",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100_GOOG_3231570.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391 (step 63328)",
            "checkpoint_step": 63328,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100_GOOG_3231570.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1-base-s63328-L100_GOOG_3231570\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 63328         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.143072+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100_GOOG_3231570.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1_26tok/checkpoints/j3162391_4zhm35uz_3162391\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1-base-s63328-L100_GOOG_3231570\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 63328         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100_GOOG_3231570.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1-base-s63328-L100_GOOG_3231570.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1-base-s63328-L100_GOOG_3231570"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100_GOOG_3231570/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100_GOOG_3231570/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1-base-s63328-L100_GOOG_3231570/data_real.tar",
            "selector": null,
            "source_value": 0.170647,
            "report_record_id": "canonical-scores:37712c45-68fd-81ca-99cb-f8f57c371a21",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ca-b1cf-f7dda2ec71b5",
      "source_row_id": "37712c45-68fd-81ca-b1cf-f7dda2ec71b5",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81cab1cff7dda2ec71b5",
      "run": "nsa-o2b-PI-FINAL",
      "stock": null,
      "model_size": null,
      "run_family": "nsa",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.616135,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.616135",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.55046,
      "reported_ks_context": 0.382344,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:14Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81cab1cff7dda2ec71b5",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:14Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:14Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:14Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.616135,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ca-b1cf-f7dda2ec71b5",
        "class": "G20",
        "model_evaluation": "nsa-o2b-PI-FINAL",
        "wws21_mean": 0.616135,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:14Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-PI-FINAL_GOOG_3260169/data_real",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-PI-FINAL_GOOG_3260169/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:14Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81cab1cff7dda2ec71b5",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 921,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922/14794",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-PI-FINAL_GOOG_3260169.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922 (step 14794)",
            "checkpoint_step": 14794,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-PI-FINAL_GOOG_3260169.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-PI-FINAL_GOOG_3260169\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.148778+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-PI-FINAL_GOOG_3260169.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2b_nsa_infer_hardened         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_O2_naive_sparse_attn/checkpoints/j3151922_6ruswt8x_3151922\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/nsa-o2b-PI-FINAL_GOOG_3260169\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 4         --checkpoint_step 14794         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-PI-FINAL_GOOG_3260169.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_nsa-o2b-PI-FINAL_GOOG_3260169.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/nsa-o2b-PI-FINAL_GOOG_3260169"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-PI-FINAL_GOOG_3260169/data_real",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-PI-FINAL_GOOG_3260169/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/nsa-o2b-PI-FINAL_GOOG_3260169/data_real",
            "selector": null,
            "source_value": 0.616135,
            "report_record_id": "canonical-scores:37712c45-68fd-81ca-b1cf-f7dda2ec71b5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81cb-946d-f88db005ad4a",
      "source_row_id": "37712c45-68fd-81cb-946d-f88db005ad4a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81cb946df88db005ad4a",
      "run": "soup-gdn-all10-rand",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.154409,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.154409",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.127062,
      "reported_ks_context": 0.078121,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:05Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81cb946df88db005ad4a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:05Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:05Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:05Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.154409,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81cb-946d-f88db005ad4a",
        "class": "G20",
        "model_evaluation": "soup-gdn-all10-rand",
        "wws21_mean": 0.154409,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:05Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-all10-rand_GOOG_2894245/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-all10-rand_GOOG_2894245/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:05Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81cb946df88db005ad4a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 922,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_all10",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-all10-rand_GOOG_2894245.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_all10 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_all10",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-all10-rand_GOOG_2894245.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_all10\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-gdn-all10-rand_GOOG_2894245\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263199+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_all10",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.182009+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_all10"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-all10-rand_GOOG_2894245.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_all10\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-gdn-all10-rand_GOOG_2894245\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-all10-rand_GOOG_2894245.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-all10-rand_GOOG_2894245.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/soup-gdn-all10-rand_GOOG_2894245"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-all10-rand_GOOG_2894245/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-all10-rand_GOOG_2894245/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-all10-rand_GOOG_2894245/data_real.tar",
            "selector": null,
            "source_value": 0.154409,
            "report_record_id": "canonical-scores:37712c45-68fd-81cb-946d-f88db005ad4a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81cb-ad3e-cba7668785f2",
      "source_row_id": "37712c45-68fd-81cb-ad3e-cba7668785f2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81cbad3ecba7668785f2",
      "run": "ctx4k-cond5500",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.148393,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.148393",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.120755,
      "reported_ks_context": 0.086384,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:08Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81cbad3ecba7668785f2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:08Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:08Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:08Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.148393,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81cb-ad3e-cba7668785f2",
        "class": "G20",
        "model_evaluation": "ctx4k-cond5500",
        "wws21_mean": 0.148393,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:08Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond5500/scores/scores_uncond_GOOG_ctx4k-cond5500_integrated_2828769.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond5500_GOOG_2828769/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:08Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond5500_GOOG_2828769/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81cbad3ecba7668785f2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 923,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond5500_GOOG_2828769/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 5500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond5500_GOOG_2828769.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond5500_GOOG_2828769\"         --n_cond_msgs 5500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond5500_GOOG_2828769.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond5500/scores/scores_uncond_GOOG_ctx4k-cond5500_integrated_2828769.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond5500/scores/scores_uncond_GOOG_ctx4k-cond5500_integrated_2828769.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14839298254337321,
        "feature_points": {
          "vol_per_min": 0.09286115669515231,
          "limit_bid_order_levels": 0.01979877513049466,
          "bid_cancellation_levels": 0.027030961263862306,
          "ask_cancellation_levels": 0.04044125016306787,
          "limit_ask_order_levels": 0.05110343277082778,
          "ask_volume_touch": 0.10556017560563205,
          "ofi_down": 0.0711260312617004,
          "ofi_up": 0.0693257873878563,
          "bid_cancellation_depth": 0.32235301176535475,
          "limit_bid_order_depth": 0.33505723501309786,
          "limit_ask_order_depth": 0.2605662299908255,
          "spread": 0.2166104994041022,
          "ask_cancellation_depth": 0.24320455955522496,
          "orderbook_imbalance": 0.11740920891998331,
          "bid_volume_touch": 0.08780490577315565,
          "ofi_stay": 0.03059992505301084,
          "ask_volume": 0.2670775980483729,
          "ofi": 0.029616292265616115,
          "log_time_to_cancel": 0.346986702968001,
          "bid_volume": 0.3465143753067767,
          "log_inter_arrival_time": 0.03520451906872169
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond5500/scores/scores_uncond_GOOG_ctx4k-cond5500_integrated_2828769.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.148393,
            "report_record_id": "canonical-scores:37712c45-68fd-81cb-ad3e-cba7668785f2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81cc-852e-c4c9e40b4e68",
      "source_row_id": "37712c45-68fd-81cc-852e-c4c9e40b4e68",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81cc852ec4c9e40b4e68",
      "run": "genlen-ctx2k-c500-g1500",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.235417,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.235417",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.208273,
      "reported_ks_context": 0.149876,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:45Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81cc852ec4c9e40b4e68",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:45Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:45Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:45Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.235417,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81cc-852e-c4c9e40b4e68",
        "class": "G20",
        "model_evaluation": "genlen-ctx2k-c500-g1500",
        "wws21_mean": 0.235417,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:45Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1500_GOOG_3405782/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1500_GOOG_3405782/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:45Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81cc852ec4c9e40b4e68",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 924,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1500_GOOG_3405782.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1500_GOOG_3405782.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g1500_GOOG_3405782\"         --n_cond_msgs 500         --n_gen_msgs 1500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263020+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.139444+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 1500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1500_GOOG_3405782.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt2k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g1500_GOOG_3405782\"         --n_cond_msgs 500         --n_gen_msgs 1500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1500_GOOG_3405782.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx2k-c500-g1500_GOOG_3405782.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx2k-c500-g1500_GOOG_3405782"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1500_GOOG_3405782/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1500_GOOG_3405782/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx2k-c500-g1500_GOOG_3405782/data_real.tar",
            "selector": null,
            "source_value": 0.235417,
            "report_record_id": "canonical-scores:37712c45-68fd-81cc-852e-c4c9e40b4e68",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81cd-9667-e6c097da143f",
      "source_row_id": "37712c45-68fd-81cd-9667-e6c097da143f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81cd9667e6c097da143f",
      "run": "phaseB-m3-14m-s500-d64-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.126248,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.126248",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.184938,
      "reported_ks_context": 0.110609,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:58Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81cd9667e6c097da143f",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:58Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:58Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:58Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.126248,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81cd-9667-e6c097da143f",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s500-d64-b-avg-c500g500",
        "wws21_mean": 0.126248,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:58Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:58Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81cd9667e6c097da143f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 925,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952/log_rank0.txt",
            "evidence_line": 576,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703790_mysq7uya_3703790",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703790_mysq7uya_3703790\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703790_mysq7uya_3703790 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952/log_rank0.txt",
                "evidence_line": 578,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790",
            "availability_checked_at": "2026-09-12T06:13:15.364479+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703790_mysq7uya_3703790\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c500g500_GOOG_3713952/data_real.tar",
            "selector": null,
            "source_value": 0.126248,
            "report_record_id": "canonical-scores:37712c45-68fd-81cd-9667-e6c097da143f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d0-a2f4-c23febc519d3",
      "source_row_id": "37712c45-68fd-81d0-a2f4-c23febc519d3",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d0a2f4c23febc519d3",
      "run": "H100_360M_step47379",
      "stock": null,
      "model_size": "360M",
      "run_family": "H100",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.182817,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.182817",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.162154,
      "reported_ks_context": 0.114778,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:39Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d0a2f4c23febc519d3",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:39Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:39Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:39Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.182817,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d0-a2f4-c23febc519d3",
        "class": "G20",
        "model_evaluation": "H100_360M_step47379",
        "wws21_mean": 0.182817,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:39Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H100_360M_step47379/scores/scores_uncond_GOOG_H100_360M_step47379_shard2_integrated_2559896.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H100_360M_step47379_GOOG_2559896/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:39Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H100_360M_step47379_GOOG_2559896/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d0a2f4c23febc519d3",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 926,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "47379",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H100-continual-pretraining/checkpoints/j2504851_1qcixnb5_2504851/47379",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H100-continual-pretraining/checkpoints/j2504851_1qcixnb5_2504851",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H100_360M_step47379_GOOG_2559896/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H100-continual-pretraining/checkpoints/j2504851_1qcixnb5_2504851/47379/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H100-continual-pretraining/checkpoints/j2504851_1qcixnb5_2504851/47379",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H100-continual-pretraining/checkpoints/j2504851_1qcixnb5_2504851/47379",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H100-continual-pretraining/checkpoints/j2504851_1qcixnb5_2504851/47379",
            "availability": "not_found",
            "path_scope": "checkpoint_step",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H100-continual-pretraining/checkpoints/j2504851_1qcixnb5_2504851/47379",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            },
            "relocation_candidates": [
              {
                "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H100-continual-pretraining_DELETING_91047/checkpoints/j2504851_1qcixnb5_2504851/47379",
                "original_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H100-continual-pretraining/checkpoints/j2504851_1qcixnb5_2504851/47379",
                "availability": "exists",
                "entries": [
                  "state",
                  "_CHECKPOINT_METADATA",
                  "metadata"
                ],
                "type": "directory",
                "path_kind": "directory",
                "note": "Possible relocated copy under _DELETING_91047; exact run and step directory exists, identity unconfirmed.",
                "identity_status": "unconfirmed_identity",
                "caveat": "Observed _DELETING_91047 directory. Exact run and step names match, but no authoritative rename record was inspected."
              }
            ]
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. An existing exact run/step candidate was found under an observed _DELETING_91047 directory. It remains a candidate because no authoritative rename record was inspected."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H100_360M_step47379_GOOG_2559896.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H100-continual-pretraining/checkpoints/j2504851_1qcixnb5_2504851\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H100_360M_step47379_GOOG_2559896\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 47379         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H100_360M_step47379_GOOG_2559896.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H100_360M_step47379/scores/scores_uncond_GOOG_H100_360M_step47379_shard2_integrated_2559896.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H100_360M_step47379/scores/scores_uncond_GOOG_H100_360M_step47379_integrated_2559896.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.18281713498475682,
        "feature_points": {
          "spread": 0.41608235486979195,
          "ask_cancellation_depth": 0.3168269402095967,
          "bid_cancellation_depth": 0.32523069646189906,
          "limit_ask_order_levels": 0.07893300942431762,
          "limit_bid_order_levels": 0.09800659748968159,
          "ask_cancellation_levels": 0.06935362702650044,
          "bid_cancellation_levels": 0.08507209378778662,
          "vol_per_min": 0.25752060768520113,
          "ofi": 0.08693791761863406,
          "ofi_up": 0.1347926773298616,
          "ofi_stay": 0.0879251869736336,
          "orderbook_imbalance": 0.11665938001482726,
          "ofi_down": 0.11054816214233637,
          "log_inter_arrival_time": 0.0875065943158836,
          "log_time_to_cancel": 0.6271192665462767,
          "ask_volume_touch": 0.055910297673708614,
          "bid_volume_touch": 0.053599576697769664,
          "ask_volume": 0.09124768340279353,
          "bid_volume": 0.07311094937967447,
          "limit_ask_order_depth": 0.32755074119223687,
          "limit_bid_order_depth": 0.33922547443748197
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H100_360M_step47379/scores/scores_uncond_GOOG_H100_360M_step47379_integrated_2559896.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.182817,
            "report_record_id": "canonical-scores:37712c45-68fd-81d0-a2f4-c23febc519d3",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d1-baec-fa1fe740a9b1",
      "source_row_id": "37712c45-68fd-81d1-baec-fa1fe740a9b1",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d1baecfa1fe740a9b1",
      "run": "phaseB-m3-14m-h128-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.199215,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.199215",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.195934,
      "reported_ks_context": 0.114168,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:07Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d1baecfa1fe740a9b1",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:07Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:07Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:07Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.199215,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d1-baec-fa1fe740a9b1",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-h128-b-avg-c250g250",
        "wws21_mean": 0.199215,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:07Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:07Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d1baecfa1fe740a9b1",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 927,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927/log_rank0.txt",
            "evidence_line": 612,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703803_39ifyzj9_3703803",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703803_39ifyzj9_3703803\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703803_39ifyzj9_3703803 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927/log_rank0.txt",
                "evidence_line": 614,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803",
            "availability_checked_at": "2026-09-12T06:13:15.371914+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703803_39ifyzj9_3703803/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703803_39ifyzj9_3703803\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-h128-b-avg-c250g250_GOOG_3713927/data_real.tar",
            "selector": null,
            "source_value": 0.199215,
            "report_record_id": "canonical-scores:37712c45-68fd-81d1-baec-fa1fe740a9b1",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d2-8128-f8d80cd04dcf",
      "source_row_id": "37712c45-68fd-81d2-8128-f8d80cd04dcf",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d28128f8d80cd04dcf",
      "run": "mamba2-77M-soup5",
      "stock": null,
      "model_size": "77M",
      "run_family": "mamba2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.143907,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.143907",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.112722,
      "reported_ks_context": 0.077481,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d28128f8d80cd04dcf",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.143907,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d2-8128-f8d80cd04dcf",
        "class": "G20",
        "model_evaluation": "mamba2-77M-soup5",
        "wws21_mean": 0.143907,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d28128f8d80cd04dcf",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 928,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5/0",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup5_GOOG_2917933.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup5_GOOG_2917933.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.182112+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup5_GOOG_2917933.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2/checkpoints/soup_mamba2_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup5_GOOG_2917933.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba2-77M-soup5_GOOG_2917933.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba2-77M-soup5_GOOG_2917933/data_real.tar",
            "selector": null,
            "source_value": 0.143907,
            "report_record_id": "canonical-scores:37712c45-68fd-81d2-8128-f8d80cd04dcf",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d2-9656-e1806154418b",
      "source_row_id": "37712c45-68fd-81d2-9656-e1806154418b",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d29656e1806154418b",
      "run": "J1-anchor1-61k-l100",
      "stock": null,
      "model_size": null,
      "run_family": "J1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.191922,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.191922",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.143315,
      "reported_ks_context": 0.105691,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:20Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d29656e1806154418b",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:20Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:20Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:20Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.191922,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d2-9656-e1806154418b",
        "class": "G20",
        "model_evaluation": "J1-anchor1-61k-l100",
        "wws21_mean": 0.191922,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:20Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor1-61k-l100/scores/scores_uncond_GOOG_J1-anchor1-61k-l100_integrated_2773503.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor1-61k-l100_GOOG_2773503/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:20Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor1-61k-l100_GOOG_2773503/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d29656e1806154418b",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 929,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "61037",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor1-61k-l100_GOOG_2773503/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=61037 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975/61037",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975/61037",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_J1-anchor1-61k-l100_GOOG_2773503.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/J1-anchor1-61k-l100_GOOG_2773503\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 61037         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_J1-anchor1-61k-l100_GOOG_2773503.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor1-61k-l100/scores/scores_uncond_GOOG_J1-anchor1-61k-l100_integrated_2773503.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor1-61k-l100/scores/scores_uncond_GOOG_J1-anchor1-61k-l100_integrated_2773503.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19192248651501528,
        "feature_points": {
          "ask_volume_touch": 0.17196729206116007,
          "ofi_down": 0.05833772345606883,
          "limit_bid_order_levels": 0.05715119864495984,
          "ask_cancellation_levels": 0.06522855605656315,
          "log_time_to_cancel": 0.5552979809591214,
          "bid_cancellation_depth": 0.3498862948448331,
          "limit_ask_order_depth": 0.30138727236398055,
          "bid_cancellation_levels": 0.05143761250491903,
          "ofi_stay": 0.04632665698321471,
          "ofi_up": 0.06681703704499126,
          "limit_ask_order_levels": 0.10279409600337747,
          "vol_per_min": 0.23377437414323293,
          "ask_volume": 0.27796945868382333,
          "spread": 0.4010223138171715,
          "orderbook_imbalance": 0.10395763184610485,
          "limit_bid_order_depth": 0.36581591431023874,
          "ask_cancellation_depth": 0.2940481738210455,
          "bid_volume": 0.28378624320161605,
          "bid_volume_touch": 0.12896928250884265,
          "ofi": 0.04499080156363106,
          "log_inter_arrival_time": 0.06940630199642549
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor1-61k-l100/scores/scores_uncond_GOOG_J1-anchor1-61k-l100_integrated_2773503.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.191922,
            "report_record_id": "canonical-scores:37712c45-68fd-81d2-9656-e1806154418b",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d3-863a-f62ed7f587ed",
      "source_row_id": "37712c45-68fd-81d3-863a-f62ed7f587ed",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d3863af62ed7f587ed",
      "run": "muon-s32155",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.203823,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.203823",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.141756,
      "reported_ks_context": 0.112263,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:05Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d3863af62ed7f587ed",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:05Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:05Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:05Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.203823,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d3-863a-f62ed7f587ed",
        "class": "G20",
        "model_evaluation": "muon-s32155",
        "wws21_mean": 0.203823,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:05Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_muon-s32155/scores/scores_uncond_GOOG_muon-s32155_integrated_2868869.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s32155_GOOG_2868869/data_gen.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:05Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s32155_GOOG_2868869/data_gen.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d3863af62ed7f587ed",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 930,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "32155",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/muon-s32155_GOOG_2868869/log_rank0.txt",
            "evidence_line": 634,
            "evidence_text": "[Checkpoint] Loading step=32155 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524/32155",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524/32155",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s32155_GOOG_2868869.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2716524_lqs66551_2716524\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/muon-s32155_GOOG_2868869\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 32155         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon-s32155_GOOG_2868869.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_muon-s32155/scores/scores_uncond_GOOG_muon-s32155_integrated_2868869.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_muon-s32155/scores/scores_uncond_GOOG_muon-s32155_integrated_2868869.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.20382254642089112,
        "feature_points": {
          "ask_cancellation_depth": 0.2970194173433549,
          "ask_cancellation_levels": 0.1429388410450134,
          "ofi_up": 0.18367184463667316,
          "ask_volume_touch": 0.13710516588159155,
          "ask_volume": 0.23822306828116677,
          "ofi_stay": 0.14879502665980193,
          "limit_bid_order_levels": 0.15298828975164347,
          "vol_per_min": 0.13602563191274,
          "bid_cancellation_levels": 0.15372934129482207,
          "limit_ask_order_levels": 0.08627765512160762,
          "ofi_down": 0.09696501579950326,
          "orderbook_imbalance": 0.09619501307500611,
          "log_time_to_cancel": 0.3984483884615548,
          "spread": 0.3961819147768618,
          "bid_cancellation_depth": 0.3350918734849049,
          "limit_bid_order_depth": 0.3452821243227857,
          "ofi": 0.14922175866357487,
          "limit_ask_order_depth": 0.31741201084937265,
          "bid_volume_touch": 0.12620262348659966,
          "bid_volume": 0.3072300713769851,
          "log_inter_arrival_time": 0.035268398613149676
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_muon-s32155/scores/scores_uncond_GOOG_muon-s32155_integrated_2868869.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.203823,
            "report_record_id": "canonical-scores:37712c45-68fd-81d3-863a-f62ed7f587ed",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d3-8b61-c8ebfc295be2",
      "source_row_id": "37712c45-68fd-81d3-8b61-c8ebfc295be2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d38b61c8ebfc295be2",
      "run": "pareto-120m-s118227",
      "stock": null,
      "model_size": "120M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.170323,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.170323",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.125853,
      "reported_ks_context": 0.095149,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:25Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d38b61c8ebfc295be2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:25Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:25Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:25Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.170323,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d3-8b61-c8ebfc295be2",
        "class": "G20",
        "model_evaluation": "pareto-120m-s118227",
        "wws21_mean": 0.170323,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:25Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s118227_GOOG_3253398/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s118227_GOOG_3253398/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:25Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d38b61c8ebfc295be2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 931,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292/118227",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s118227_GOOG_3253398.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292 (step 118227)",
            "checkpoint_step": 118227,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s118227_GOOG_3253398.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s118227_GOOG_3253398\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 118227         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.165112+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s118227_GOOG_3253398.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2505292_7dbjibmu_2505292\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s118227_GOOG_3253398\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 118227         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s118227_GOOG_3253398.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-120m-s118227_GOOG_3253398.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-120m-s118227_GOOG_3253398"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s118227_GOOG_3253398/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s118227_GOOG_3253398/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-120m-s118227_GOOG_3253398/data_real.tar",
            "selector": null,
            "source_value": 0.170323,
            "report_record_id": "canonical-scores:37712c45-68fd-81d3-8b61-c8ebfc295be2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d3-b144-d25e4eff36a2",
      "source_row_id": "37712c45-68fd-81d3-b144-d25e4eff36a2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d3b144d25e4eff36a2",
      "run": "H2_55M_ctx4k_step12458",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.246579,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.246579",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.188197,
      "reported_ks_context": 0.129007,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:09Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d3b144d25e4eff36a2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:09Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:09Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:09Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.246579,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d3-b144-d25e4eff36a2",
        "class": "G20",
        "model_evaluation": "H2_55M_ctx4k_step12458",
        "wws21_mean": 0.246579,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:09Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2_55M_ctx4k_step12458/scores/scores_uncond_GOOG_H2_55M_ctx4k_step12458_shard9_integrated_2566031.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2_55M_ctx4k_step12458_GOOG_2566031/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:09Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2_55M_ctx4k_step12458_GOOG_2566031/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d3b144d25e4eff36a2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 932,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "12458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504026_ss7hmq26_2504026/12458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504026_ss7hmq26_2504026",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2_55M_ctx4k_step12458_GOOG_2566031/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504026_ss7hmq26_2504026/12458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504026_ss7hmq26_2504026/12458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504026_ss7hmq26_2504026/12458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "12458",
                "15033",
                "17605",
                "20180",
                "9885",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504026_ss7hmq26_2504026/12458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2_55M_ctx4k_step12458_GOOG_2566031.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504026_ss7hmq26_2504026\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2_55M_ctx4k_step12458_GOOG_2566031\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 12458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2_55M_ctx4k_step12458_GOOG_2566031.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2_55M_ctx4k_step12458/scores/scores_uncond_GOOG_H2_55M_ctx4k_step12458_shard9_integrated_2566031.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2_55M_ctx4k_step12458/scores/scores_uncond_GOOG_H2_55M_ctx4k_step12458_integrated_2566031.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.24657908228865072,
        "feature_points": {
          "spread": 0.4872373923479694,
          "ask_cancellation_depth": 0.38682879169488665,
          "bid_cancellation_depth": 0.41315382405324474,
          "limit_ask_order_levels": 0.09813089451471911,
          "limit_bid_order_levels": 0.11899269422843645,
          "ask_cancellation_levels": 0.1713826762785146,
          "bid_cancellation_levels": 0.12263982937400446,
          "vol_per_min": 0.13430021518767704,
          "ofi": 0.23809927956879678,
          "ofi_up": 0.27901179928780684,
          "ofi_stay": 0.23772511101176566,
          "orderbook_imbalance": 0.12106570511000053,
          "ofi_down": 0.22315777148667293,
          "log_inter_arrival_time": 0.0703081283303887,
          "log_time_to_cancel": 0.5048037321662753,
          "ask_volume_touch": 0.20652913800224557,
          "bid_volume_touch": 0.21239036674618503,
          "ask_volume": 0.136761306111497,
          "bid_volume": 0.21932322500703993,
          "limit_ask_order_depth": 0.40459150258024085,
          "limit_bid_order_depth": 0.39172734497329764
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2_55M_ctx4k_step12458/scores/scores_uncond_GOOG_H2_55M_ctx4k_step12458_integrated_2566031.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.246579,
            "report_record_id": "canonical-scores:37712c45-68fd-81d3-b144-d25e4eff36a2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d4-85cc-e254e3d0aec6",
      "source_row_id": "37712c45-68fd-81d4-85cc-e254e3d0aec6",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d485cce254e3d0aec6",
      "run": "R1-s4k-d256-raw-s48210",
      "stock": null,
      "model_size": null,
      "run_family": "R1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.124882,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.124882",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.171014,
      "reported_ks_context": 0.094091,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:25Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d485cce254e3d0aec6",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:25Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:25Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:25Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.124882,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d4-85cc-e254e3d0aec6",
        "class": "G20",
        "model_evaluation": "R1-s4k-d256-raw-s48210",
        "wws21_mean": 0.124882,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:25Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s48210_GOOG_3745462/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s48210_GOOG_3745462/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:25Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d485cce254e3d0aec6",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 933,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096/48210",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s48210_GOOG_3745462.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096 (step 48210)",
            "checkpoint_step": 48210,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s48210_GOOG_3745462.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s48210_GOOG_3745462\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 48210         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.136864+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s48210_GOOG_3745462.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s48210_GOOG_3745462\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 48210         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s48210_GOOG_3745462.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s48210_GOOG_3745462.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s48210_GOOG_3745462"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s48210_GOOG_3745462/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s48210_GOOG_3745462/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s48210_GOOG_3745462/data_real.tar",
            "selector": null,
            "source_value": 0.124882,
            "report_record_id": "canonical-scores:37712c45-68fd-81d4-85cc-e254e3d0aec6",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d5-8412-f307f665327d",
      "source_row_id": "37712c45-68fd-81d5-8412-f307f665327d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d58412f307f665327d",
      "run": "muon_j2727131_52619",
      "stock": null,
      "model_size": null,
      "run_family": "muon",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.160786,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.160786",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.13026,
      "reported_ks_context": 0.089809,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-11T06:58:04Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2047",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-11T06:58:04Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-11T06:58:04Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-11T06:58:04Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.160786,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d5-8412-f307f665327d",
        "class": "G20",
        "model_evaluation": "muon_j2727131_52619",
        "wws21_mean": 0.160786,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-11T06:58:04Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_muon_j2727131_52619/scores/scores_uncond_GOOG_muon_j2727131_52619_integrated_2773471.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/muon_j2727131_52619_GOOG_2773471/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-11T06:58:04Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/muon_j2727131_52619_GOOG_2773471/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d58412f307f665327d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 934,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "52619",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/muon_j2727131_52619_GOOG_2773471/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=52619 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131/52619",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131/52619",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon_j2727131_52619_GOOG_2773471.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2727131_d6gdaz35_2727131\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/muon_j2727131_52619_GOOG_2773471\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 52619         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_muon_j2727131_52619_GOOG_2773471.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_muon_j2727131_52619/scores/scores_uncond_GOOG_muon_j2727131_52619_integrated_2773471.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_muon_j2727131_52619/scores/scores_uncond_GOOG_muon_j2727131_52619_integrated_2773471.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.16078627807935858,
        "feature_points": {
          "limit_bid_order_levels": 0.059006540160635834,
          "ofi_down": 0.07239728159869648,
          "bid_volume": 0.2066020957734242,
          "log_inter_arrival_time": 0.03743080560552561,
          "ofi_stay": 0.09693634874449467,
          "ask_cancellation_levels": 0.04622045639299696,
          "bid_cancellation_levels": 0.025653388896841746,
          "limit_ask_order_levels": 0.04660741568290203,
          "bid_cancellation_depth": 0.3325886613387905,
          "bid_volume_touch": 0.05744121297299073,
          "ask_volume": 0.12281326252225296,
          "ofi_up": 0.11514973800784944,
          "vol_per_min": 0.09863198699636228,
          "spread": 0.37632631977215936,
          "ask_cancellation_depth": 0.30389176825953634,
          "limit_ask_order_depth": 0.32255226997409076,
          "limit_bid_order_depth": 0.3419070218315702,
          "ask_volume_touch": 0.046867044599662906,
          "orderbook_imbalance": 0.10581808368866544,
          "log_time_to_cancel": 0.46479791527955167,
          "ofi": 0.09687222156753014
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_muon_j2727131_52619/scores/scores_uncond_GOOG_muon_j2727131_52619_integrated_2773471.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.160786,
            "report_record_id": "canonical-scores:37712c45-68fd-81d5-8412-f307f665327d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d5-b5e5-e042eb7a6627",
      "source_row_id": "37712c45-68fd-81d5-b5e5-e042eb7a6627",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d5b5e5e042eb7a6627",
      "run": "adamw-s19969",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.188681,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.188681",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.14174,
      "reported_ks_context": 0.10421,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:36Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d5b5e5e042eb7a6627",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:36Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:36Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:36Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.188681,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d5-b5e5-e042eb7a6627",
        "class": "G20",
        "model_evaluation": "adamw-s19969",
        "wws21_mean": 0.188681,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:36Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s19969_GOOG_2868483/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s19969_GOOG_2868483/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:36Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d5b5e5e042eb7a6627",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 935,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s19969_GOOG_2868483.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747 (step 19969)",
            "checkpoint_step": 19969,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s19969_GOOG_2868483.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s19969_GOOG_2868483\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 19969         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263082+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158732+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s19969_GOOG_2868483.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s19969_GOOG_2868483\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 19969         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s19969_GOOG_2868483.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s19969_GOOG_2868483.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-s19969_GOOG_2868483"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s19969_GOOG_2868483/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s19969_GOOG_2868483/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s19969_GOOG_2868483/data_real.tar",
            "selector": null,
            "source_value": 0.188681,
            "report_record_id": "canonical-scores:37712c45-68fd-81d5-b5e5-e042eb7a6627",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d6-976a-cbae08873a18",
      "source_row_id": "37712c45-68fd-81d6-976a-cbae08873a18",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d6976acbae08873a18",
      "run": "j2504167_step100378",
      "stock": null,
      "model_size": null,
      "run_family": "j2504167",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.153932,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.153932",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.123699,
      "reported_ks_context": 0.084534,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T21:58:46Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2180",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T21:58:46Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T21:58:46Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T21:58:46Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.153932,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d6-976a-cbae08873a18",
        "class": "G20",
        "model_evaluation": "j2504167_step100378",
        "wws21_mean": 0.153932,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T21:58:46Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167_step100378/scores/scores_uncond_GOOG_j2504167_step100378_shard3_integrated_2541166.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167_step100378_GOOG_2541166/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T21:58:46Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167_step100378_GOOG_2541166/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d6976acbae08873a18",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 936,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504167_step100378_GOOG_2541166/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504167_step100378_GOOG_2541166.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504167_step100378_GOOG_2541166\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504167_step100378_GOOG_2541166.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167_step100378/scores/scores_uncond_GOOG_j2504167_step100378_shard3_integrated_2541166.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167_step100378/scores/scores_uncond_GOOG_j2504167_step100378_integrated_2541166.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15393164371708865,
        "feature_points": {
          "spread": 0.204515672057952,
          "ask_cancellation_depth": 0.3284057950294012,
          "bid_cancellation_depth": 0.36083599258449683,
          "limit_ask_order_levels": 0.04204001009506034,
          "limit_bid_order_levels": 0.050718019896852595,
          "ask_cancellation_levels": 0.06656149292243056,
          "bid_cancellation_levels": 0.07085140711145414,
          "vol_per_min": 0.1575581193845981,
          "ofi": 0.0907628178624395,
          "ofi_up": 0.08901507147102966,
          "ofi_stay": 0.09099773431614548,
          "orderbook_imbalance": 0.09990896648847886,
          "ofi_down": 0.05790478806274494,
          "log_inter_arrival_time": 0.05530026412765801,
          "log_time_to_cancel": 0.39434523018889117,
          "ask_volume_touch": 0.07395047748275527,
          "bid_volume_touch": 0.10413158261880229,
          "ask_volume": 0.07908566489153424,
          "bid_volume": 0.10771188810219112,
          "limit_ask_order_depth": 0.3477057773659427,
          "limit_bid_order_depth": 0.3602577459980026
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504167_step100378/scores/scores_uncond_GOOG_j2504167_step100378_integrated_2541166.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.153932,
            "report_record_id": "canonical-scores:37712c45-68fd-81d6-976a-cbae08873a18",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d6-aace-c3031725a530",
      "source_row_id": "37712c45-68fd-81d6-aace-c3031725a530",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d6aacec3031725a530",
      "run": "K1-GDN-j2722414-convfix",
      "stock": null,
      "model_size": null,
      "run_family": "K1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.147222,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.147222",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.119967,
      "reported_ks_context": 0.075878,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-11T02:56:19Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2057",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-11T02:56:19Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-11T02:56:19Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-11T02:56:19Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.147222,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d6-aace-c3031725a530",
        "class": "G20",
        "model_evaluation": "K1-GDN-j2722414-convfix",
        "wws21_mean": 0.147222,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-11T02:56:19Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_K1-GDN-j2722414-convfix/scores/scores_uncond_GOOG_K1-GDN-j2722414-convfix_integrated_2824117.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/K1-GDN-j2722414-convfix_GOOG_2824117/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-11T02:56:19Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/K1-GDN-j2722414-convfix_GOOG_2824117/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d6aacec3031725a530",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 937,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "23356",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/K1-GDN-j2722414-convfix_GOOG_2824117/log_rank0.txt",
            "evidence_line": 752,
            "evidence_text": "[Checkpoint] Loading step=23356 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414/23356",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "10850",
                "12413",
                "13978",
                "15545",
                "17111",
                "18675",
                "20237",
                "21799",
                "23356",
                "9287",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414/23356",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_K1-GDN-j2722414-convfix_GOOG_2824117.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K1_GDN/checkpoints/j2722414_mr8w0rjg_2722414\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/K1-GDN-j2722414-convfix_GOOG_2824117\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 23356         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_K1-GDN-j2722414-convfix_GOOG_2824117.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_K1-GDN-j2722414-convfix/scores/scores_uncond_GOOG_K1-GDN-j2722414-convfix_integrated_2824117.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_K1-GDN-j2722414-convfix/scores/scores_uncond_GOOG_K1-GDN-j2722414-convfix_integrated_2824117.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1472222511078364,
        "feature_points": {
          "ofi_up": 0.08469333644733523,
          "vol_per_min": 0.0739617110391497,
          "ask_volume_touch": 0.07814917653441729,
          "bid_volume": 0.1293893110986038,
          "ask_cancellation_depth": 0.33833651233250844,
          "limit_bid_order_levels": 0.02426956968828853,
          "bid_volume_touch": 0.071660512142713,
          "bid_cancellation_depth": 0.3932878906374001,
          "limit_ask_order_depth": 0.35039722230585696,
          "ofi_down": 0.11865420840743948,
          "spread": 0.12174158750732797,
          "limit_ask_order_levels": 0.03163480157341928,
          "ask_cancellation_levels": 0.03538694243629965,
          "ofi_stay": 0.13896862981941016,
          "bid_cancellation_levels": 0.029255901940843137,
          "log_inter_arrival_time": 0.08026884261395112,
          "ofi": 0.13820534140177818,
          "limit_bid_order_depth": 0.39857587632547875,
          "orderbook_imbalance": 0.09466002693142878,
          "ask_volume": 0.12219860055071094,
          "log_time_to_cancel": 0.23797127153020387
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_K1-GDN-j2722414-convfix/scores/scores_uncond_GOOG_K1-GDN-j2722414-convfix_integrated_2824117.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.147222,
            "report_record_id": "canonical-scores:37712c45-68fd-81d6-aace-c3031725a530",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d6-b5b6-e9b1eef1e3e2",
      "source_row_id": "37712c45-68fd-81d6-b5b6-e9b1eef1e3e2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d6b5b6e9b1eef1e3e2",
      "run": "pareto-360m-muon-s12806",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.19623,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.19623",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.1585,
      "reported_ks_context": 0.102158,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:34Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d6b5b6e9b1eef1e3e2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:34Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:34Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:34Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.19623,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d6-b5b6-e9b1eef1e3e2",
        "class": "G20",
        "model_evaluation": "pareto-360m-muon-s12806",
        "wws21_mean": 0.19623,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:34Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s12806_GOOG_3253411/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s12806_GOOG_3253411/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:34Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d6b5b6e9b1eef1e3e2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 938,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s12806_GOOG_3253411.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592 (step 12806)",
            "checkpoint_step": 12806,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s12806_GOOG_3253411.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s12806_GOOG_3253411\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 12806         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263086+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158865+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s12806_GOOG_3253411.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2878592_35aavknm_2878592\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s12806_GOOG_3253411\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 12806         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s12806_GOOG_3253411.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-muon-s12806_GOOG_3253411.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-muon-s12806_GOOG_3253411"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s12806_GOOG_3253411/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s12806_GOOG_3253411/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-muon-s12806_GOOG_3253411/data_real.tar",
            "selector": null,
            "source_value": 0.19623,
            "report_record_id": "canonical-scores:37712c45-68fd-81d6-b5b6-e9b1eef1e3e2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81d8-871d-d38578e1831c",
      "source_row_id": "37712c45-68fd-81d8-871d-d38578e1831c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81d8871dd38578e1831c",
      "run": "s4k-d512-32n-step22400",
      "stock": null,
      "model_size": null,
      "run_family": "s4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.094397,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.094397",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.161156,
      "reported_ks_context": 0.09878,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:12Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81d8871dd38578e1831c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:12Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:12Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:12Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.094397,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81d8-871d-d38578e1831c",
        "class": "G20",
        "model_evaluation": "s4k-d512-32n-step22400",
        "wws21_mean": 0.094397,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:12Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step22400_GOOG_3763546/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step22400_GOOG_3763546/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:12Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81d8871dd38578e1831c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 939,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036/22400",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step22400_GOOG_3763546.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036 (step 22400)",
            "checkpoint_step": 22400,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step22400_GOOG_3763546.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step22400_GOOG_3763546\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 22400         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.176082+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step22400_GOOG_3763546.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step22400_GOOG_3763546\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 22400         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step22400_GOOG_3763546.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step22400_GOOG_3763546.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step22400_GOOG_3763546"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step22400_GOOG_3763546/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step22400_GOOG_3763546/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step22400_GOOG_3763546/data_real.tar",
            "selector": null,
            "source_value": 0.094397,
            "report_record_id": "canonical-scores:37712c45-68fd-81d8-871d-d38578e1831c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81da-9065-dac9c711938c",
      "source_row_id": "37712c45-68fd-81da-9065-dac9c711938c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81da9065dac9c711938c",
      "run": "j2504811_hfmn2548",
      "stock": null,
      "model_size": null,
      "run_family": "j2504811",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.231756,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.231756",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.189904,
      "reported_ks_context": 0.119551,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T22:00:39Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2174",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T22:00:39Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T22:00:39Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T22:00:39Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.231756,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81da-9065-dac9c711938c",
        "class": "G20",
        "model_evaluation": "j2504811_hfmn2548",
        "wws21_mean": 0.231756,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T22:00:39Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_shard7_integrated_2566546.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546/data_real.tar;/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566544/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T22:00:39Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81da9065dac9c711938c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 940,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "192987",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546/log_rank0.txt",
            "evidence_line": 628,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "103505",
                "108205",
                "112915",
                "160000",
                "164710",
                "174140",
                "178853",
                "183557",
                "188270",
                "192987",
                "197705",
                "202424",
                "207134",
                "211843",
                "216558",
                "23310",
                "28025",
                "32740",
                "37455",
                "56320",
                "70465",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811/192987",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Multiple source input candidates were resolved by the exact score filename explicitly listed in the matching integrated-job log. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 520,
            "text": "  - /projects/s5e/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_shard7_integrated_2566546.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H200-fuse-token/checkpoints/j2504811_hfmn2548_2504811\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504811_hfmn2548_GOOG_2566546\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 192987         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 12,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504811_hfmn2548_GOOG_2566546.out",
            "line": 520,
            "text": "  - /projects/s5e/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_shard7_integrated_2566546.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_shard7_integrated_2566546.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_integrated_2566546.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.23175622688048572,
        "feature_points": {
          "spread": 0.4228574483487797,
          "ask_cancellation_depth": 0.450635592914468,
          "bid_cancellation_depth": 0.472347188503442,
          "limit_ask_order_levels": 0.1174546570881266,
          "limit_bid_order_levels": 0.10914834105981358,
          "ask_cancellation_levels": 0.16848615565439748,
          "bid_cancellation_levels": 0.13567124666286237,
          "vol_per_min": 0.13371054418951422,
          "ofi": 0.11802831948540084,
          "ofi_up": 0.15313352338481845,
          "ofi_stay": 0.12045877340386857,
          "orderbook_imbalance": 0.11082441992647245,
          "ofi_down": 0.10881454275305076,
          "log_inter_arrival_time": 0.20927945307560433,
          "log_time_to_cancel": 0.6670623874738916,
          "ask_volume_touch": 0.16926273641660505,
          "bid_volume_touch": 0.10257804957957822,
          "ask_volume": 0.11102227410808888,
          "bid_volume": 0.07750528565099786,
          "limit_ask_order_depth": 0.4521070202383495,
          "limit_bid_order_depth": 0.4564928045720696
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504811_hfmn2548/scores/scores_uncond_GOOG_j2504811_hfmn2548_integrated_2566546.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.231756,
            "report_record_id": "canonical-scores:37712c45-68fd-81da-9065-dac9c711938c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81da-918d-c47b57c95a7b",
      "source_row_id": "37712c45-68fd-81da-918d-c47b57c95a7b",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81da918dc47b57c95a7b",
      "run": "ticks-ctx4k-rand",
      "stock": null,
      "model_size": null,
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.208224,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.208224",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.160006,
      "reported_ks_context": 0.106359,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:29Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81da918dc47b57c95a7b",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:29Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:29Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:29Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.208224,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81da-918d-c47b57c95a7b",
        "class": "G20",
        "model_evaluation": "ticks-ctx4k-rand",
        "wws21_mean": 0.208224,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:29Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx4k-rand/scores/scores_uncond_GOOG_ticks-ctx4k-rand_integrated_3278230.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:29Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81da918dc47b57c95a7b",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 941,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144/log_rank0.txt",
            "evidence_line": 634,
            "evidence_text": "[Checkpoint] Loading step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx4k-rand_GOOG_3278230.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx4k-rand_GOOG_3278230.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx4k-rand_GOOG_3278230.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-ctx4k-rand_integrated_3278230.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt4k-rand_GOOG_2888144.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt4k-rand_GOOG_2888144.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx4k-rand_GOOG_3278230.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx4k-rand_GOOG_3278230.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-ctx4k-rand_GOOG_3278230.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-ctx4k-rand_integrated_3278230.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx4k-rand/scores/scores_uncond_GOOG_ticks-ctx4k-rand_integrated_3278230.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx4k-rand/scores/scores_uncond_GOOG_ticks-ctx4k-rand_integrated_3278230.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.20822429687201505,
        "feature_points": {
          "ask_volume_touch": 0.07423113948036895,
          "ask_cancellation_depth": 0.32214696886997973,
          "ofi_up": 0.026788829561577708,
          "bid_volume_touch": 0.06641045761246984,
          "ask_cancellation_ticks": 0.32122868479052474,
          "limit_ask_order_depth": 0.33361999603179837,
          "ask_volume": 0.19108755379435402,
          "ofi": 0.05184401697676723,
          "limit_bid_order_depth": 0.36450556248818666,
          "log_inter_arrival_time": 0.05089693047507562,
          "ofi_down": 0.033934704965238656,
          "vol_per_min": 0.09140839860164503,
          "ofi_stay": 0.05274383715473822,
          "spread": 0.27214322604454844,
          "limit_bid_order_ticks": 0.3644825863725359,
          "limit_ask_order_ticks": 0.3335973578331047,
          "bid_cancellation_depth": 0.3582993776552169,
          "bid_cancellation_ticks": 0.35735231691738445,
          "log_time_to_cancel": 0.3723300140596261,
          "orderbook_imbalance": 0.10937980308180936,
          "bid_volume": 0.22427847154536512
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-ctx4k-rand/scores/scores_uncond_GOOG_ticks-ctx4k-rand_integrated_3278230.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.208224,
            "report_record_id": "canonical-scores:37712c45-68fd-81da-918d-c47b57c95a7b",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81db-8057-ce0cea2e6391",
      "source_row_id": "37712c45-68fd-81db-8057-ce0cea2e6391",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81db8057ce0cea2e6391",
      "run": "s4k-d256-s55170",
      "stock": null,
      "model_size": null,
      "run_family": "s4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.104277,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.104277",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.148578,
      "reported_ks_context": 0.09123,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:44Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81db8057ce0cea2e6391",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:44Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:44Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:44Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.104277,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81db-8057-ce0cea2e6391",
        "class": "G20",
        "model_evaluation": "s4k-d256-s55170",
        "wws21_mean": 0.104277,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:44Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s55170_GOOG_3848789/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s55170_GOOG_3848789/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:44Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81db8057ce0cea2e6391",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 942,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699/55170",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s55170_GOOG_3848789.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699 (step 55170)",
            "checkpoint_step": 55170,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s55170_GOOG_3848789.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d256-s55170_GOOG_3848789\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 55170         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.171015+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s55170_GOOG_3848789.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3825699_kzrc5cpg_3825699\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d256-s55170_GOOG_3848789\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 55170         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s55170_GOOG_3848789.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d256-s55170_GOOG_3848789.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s4k-d256-s55170_GOOG_3848789"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s55170_GOOG_3848789/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s55170_GOOG_3848789/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d256-s55170_GOOG_3848789/data_real.tar",
            "selector": null,
            "source_value": 0.104277,
            "report_record_id": "canonical-scores:37712c45-68fd-81db-8057-ce0cea2e6391",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81db-84f2-f301e571bf13",
      "source_row_id": "37712c45-68fd-81db-84f2-f301e571bf13",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81db84f2f301e571bf13",
      "run": "pareto-360m-adamw-s60462",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.185693,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.185693",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.142357,
      "reported_ks_context": 0.101937,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:33Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81db84f2f301e571bf13",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:33Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:33Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:33Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.185693,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81db-84f2-f301e571bf13",
        "class": "G20",
        "model_evaluation": "pareto-360m-adamw-s60462",
        "wws21_mean": 0.185693,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:33Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s60462_GOOG_3253404/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s60462_GOOG_3253404/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:33Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81db84f2f301e571bf13",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 943,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s60462_GOOG_3253404.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665 (step 60462)",
            "checkpoint_step": 60462,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s60462_GOOG_3253404.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s60462_GOOG_3253404\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 60462         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262928+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.121688+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s60462_GOOG_3253404.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s60462_GOOG_3253404\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 60462         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s60462_GOOG_3253404.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s60462_GOOG_3253404.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s60462_GOOG_3253404"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s60462_GOOG_3253404/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s60462_GOOG_3253404/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s60462_GOOG_3253404/data_real.tar",
            "selector": null,
            "source_value": 0.185693,
            "report_record_id": "canonical-scores:37712c45-68fd-81db-84f2-f301e571bf13",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81db-899e-ee0ea18e06da",
      "source_row_id": "37712c45-68fd-81db-899e-ee0ea18e06da",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81db899eee0ea18e06da",
      "run": "linf-26tok-split-55M",
      "stock": null,
      "model_size": "55M",
      "run_family": "linf",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.22762,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.22762",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.201099,
      "reported_ks_context": 0.151091,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:55Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81db899eee0ea18e06da",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:55Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:55Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:55Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.22762,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81db-899e-ee0ea18e06da",
        "class": "G20",
        "model_evaluation": "linf-26tok-split-55M",
        "wws21_mean": 0.22762,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:55Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/linf-26tok-split-55M_GOOG_3249756/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/linf-26tok-split-55M_GOOG_3249756/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:55Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81db899eee0ea18e06da",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 944,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_linf-26tok-split-55M_GOOG_3249756.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129 (step 62183)",
            "checkpoint_step": 62183,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_linf-26tok-split-55M_GOOG_3249756.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/linf-26tok-split-55M_GOOG_3249756\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 16         --checkpoint_step 62183         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_linf/GOOG\" --wide_levels 7500 --gt_compare"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263186+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.181399+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_linf-26tok-split-55M_GOOG_3249756.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/linf-26tok-split-55M_GOOG_3249756\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 16         --checkpoint_step 62183         --test_split 1.0 --sample_indices_file \"/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_linf/GOOG\" --wide_levels 7500 --gt_compare"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_linf-26tok-split-55M_GOOG_3249756.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_linf-26tok-split-55M_GOOG_3249756.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/linf-26tok-split-55M_GOOG_3249756"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/linf-26tok-split-55M_GOOG_3249756/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/linf-26tok-split-55M_GOOG_3249756/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/linf-26tok-split-55M_GOOG_3249756/data_real.tar",
            "selector": null,
            "source_value": 0.22762,
            "report_record_id": "canonical-scores:37712c45-68fd-81db-899e-ee0ea18e06da",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81db-bd67-c412801c538c",
      "source_row_id": "37712c45-68fd-81db-bd67-c412801c538c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81dbbd67c412801c538c",
      "run": "adamw-32n-s29069",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.148016,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.148016",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.115525,
      "reported_ks_context": 0.082789,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:02Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81dbbd67c412801c538c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:02Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:02Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:02Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.148016,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81db-bd67-c412801c538c",
        "class": "G20",
        "model_evaluation": "adamw-32n-s29069",
        "wws21_mean": 0.148016,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:02Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29069_GOOG_2871388/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29069_GOOG_2871388/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:02Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81dbbd67c412801c538c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 945,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29069_GOOG_2871388.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 29069)",
            "checkpoint_step": 29069,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29069_GOOG_2871388.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s29069_GOOG_2871388\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 29069         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29069_GOOG_2871388.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s29069_GOOG_2871388\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 29069         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29069_GOOG_2871388.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29069_GOOG_2871388.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s29069_GOOG_2871388"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29069_GOOG_2871388/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29069_GOOG_2871388/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29069_GOOG_2871388/data_real.tar",
            "selector": null,
            "source_value": 0.148016,
            "report_record_id": "canonical-scores:37712c45-68fd-81db-bd67-c412801c538c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81dc-af12-f6f6afbf8a28",
      "source_row_id": "37712c45-68fd-81dc-af12-f6f6afbf8a28",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81dcaf12f6f6afbf8a28",
      "run": "adamw-s60462",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.165353,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.165353",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.131247,
      "reported_ks_context": 0.085092,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:56Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81dcaf12f6f6afbf8a28",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:56Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:56Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:56Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.165353,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81dc-af12-f6f6afbf8a28",
        "class": "G20",
        "model_evaluation": "adamw-s60462",
        "wws21_mean": 0.165353,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:56Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s60462_GOOG_2868486/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s60462_GOOG_2868486/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:56Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81dcaf12f6f6afbf8a28",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 946,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s60462_GOOG_2868486.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665 (step 60462)",
            "checkpoint_step": 60462,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s60462_GOOG_2868486.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s60462_GOOG_2868486\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 60462         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262928+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.121688+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s60462_GOOG_2868486.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2715665_ueew9sq2_2715665\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-s60462_GOOG_2868486\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 60462         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s60462_GOOG_2868486.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-s60462_GOOG_2868486.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-s60462_GOOG_2868486"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s60462_GOOG_2868486/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s60462_GOOG_2868486/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-s60462_GOOG_2868486/data_real.tar",
            "selector": null,
            "source_value": 0.165353,
            "report_record_id": "canonical-scores:37712c45-68fd-81dc-af12-f6f6afbf8a28",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81de-ac67-d1db86a35b2d",
      "source_row_id": "37712c45-68fd-81de-ac67-d1db86a35b2d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81deac67d1db86a35b2d",
      "run": "pareto-360m-adamw-s22789",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.246353,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.246353",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.190219,
      "reported_ks_context": 0.135805,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:56Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81deac67d1db86a35b2d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:56Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:56Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:56Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.246353,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81de-ac67-d1db86a35b2d",
        "class": "G20",
        "model_evaluation": "pareto-360m-adamw-s22789",
        "wws21_mean": 0.246353,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:56Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s22789_GOOG_3253402/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s22789_GOOG_3253402/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:56Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81deac67d1db86a35b2d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 947,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s22789_GOOG_3253402.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747 (step 22789)",
            "checkpoint_step": 22789,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s22789_GOOG_3253402.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s22789_GOOG_3253402\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 22789         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263082+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158732+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s22789_GOOG_3253402.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s22789_GOOG_3253402\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 22789         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s22789_GOOG_3253402.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s22789_GOOG_3253402.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s22789_GOOG_3253402"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s22789_GOOG_3253402/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s22789_GOOG_3253402/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s22789_GOOG_3253402/data_real.tar",
            "selector": null,
            "source_value": 0.246353,
            "report_record_id": "canonical-scores:37712c45-68fd-81de-ac67-d1db86a35b2d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81de-af76-d7c6c36886ce",
      "source_row_id": "37712c45-68fd-81de-af76-d7c6c36886ce",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81deaf76d7c6c36886ce",
      "run": "phaseA-r075-a-s30960-c250g250",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.131515,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.131515",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.159438,
      "reported_ks_context": 0.105224,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:07Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81deaf76d7c6c36886ce",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:07Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:07Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:07Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.131515,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81de-af76-d7c6c36886ce",
        "class": "G20",
        "model_evaluation": "phaseA-r075-a-s30960-c250g250",
        "wws21_mean": 0.131515,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:07Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c250g250_GOOG_3701970/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c250g250_GOOG_3701970/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:07Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81deaf76d7c6c36886ce",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 948,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347/30960",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c250g250_GOOG_3701970.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347 (step 30960)",
            "checkpoint_step": 30960,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c250g250_GOOG_3701970.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r075-a-s30960-c250g250_GOOG_3701970\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 30960         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.141730+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c250g250_GOOG_3701970.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696347_qaf58ioi_3696347\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r075-a-s30960-c250g250_GOOG_3701970\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 30960         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c250g250_GOOG_3701970.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r075-a-s30960-c250g250_GOOG_3701970.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-r075-a-s30960-c250g250_GOOG_3701970"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c250g250_GOOG_3701970/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c250g250_GOOG_3701970/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r075-a-s30960-c250g250_GOOG_3701970/data_real.tar",
            "selector": null,
            "source_value": 0.131515,
            "report_record_id": "canonical-scores:37712c45-68fd-81de-af76-d7c6c36886ce",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81df-8d4b-d3a1a1bd6dcf",
      "source_row_id": "37712c45-68fd-81df-8d4b-d3a1a1bd6dcf",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81df8d4bd3a1a1bd6dcf",
      "run": "46m-s2k-d256-step26312",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.129802,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.129802",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.177459,
      "reported_ks_context": 0.111045,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:36:24Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81df8d4bd3a1a1bd6dcf",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:36:24Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:36:24Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:36:24Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.129802,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81df-8d4b-d3a1a1bd6dcf",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-step26312",
        "wws21_mean": 0.129802,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:36:24Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-step26312_GOOG_3775011/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-step26312_GOOG_3775011/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:36:24Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81df8d4bd3a1a1bd6dcf",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 949,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763899_3t4zhru3_3763899/26312",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-step26312_GOOG_3775011.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763899_3t4zhru3_3763899 (step 26312)",
            "checkpoint_step": 26312,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763899_3t4zhru3_3763899",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-step26312_GOOG_3775011.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763899_3t4zhru3_3763899\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-step26312_GOOG_3775011\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 26312         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.182362+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763899_3t4zhru3_3763899",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763899_3t4zhru3_3763899"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-step26312_GOOG_3775011.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3763899_3t4zhru3_3763899\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-step26312_GOOG_3775011\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 26312         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-step26312_GOOG_3775011.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-step26312_GOOG_3775011.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-step26312_GOOG_3775011"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-step26312_GOOG_3775011/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-step26312_GOOG_3775011/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-step26312_GOOG_3775011/data_real.tar",
            "selector": null,
            "source_value": 0.129802,
            "report_record_id": "canonical-scores:37712c45-68fd-81df-8d4b-d3a1a1bd6dcf",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81df-90ea-e73694dfcb75",
      "source_row_id": "37712c45-68fd-81df-90ea-e73694dfcb75",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81df90eae73694dfcb75",
      "run": "s2k-d256-match14m-s37050-r2",
      "stock": null,
      "model_size": "14M",
      "run_family": "s2k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.122805,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.122805",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.165695,
      "reported_ks_context": 0.095488,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:36Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81df90eae73694dfcb75",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:36Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:36Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:36Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.122805,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81df-90ea-e73694dfcb75",
        "class": "G20",
        "model_evaluation": "s2k-d256-match14m-s37050-r2",
        "wws21_mean": 0.122805,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:36Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-match14m-s37050-r2_GOOG_3833656/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-match14m-s37050-r2_GOOG_3833656/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:36Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81df90eae73694dfcb75",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 950,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596/37050",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-match14m-s37050-r2_GOOG_3833656.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596 (step 37050)",
            "checkpoint_step": 37050,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-match14m-s37050-r2_GOOG_3833656.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-match14m-s37050-r2_GOOG_3833656\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 37050         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.121699+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-match14m-s37050-r2_GOOG_3833656.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3817596_04x4mrx2_3817596\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s2k-d256-match14m-s37050-r2_GOOG_3833656\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 37050         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-match14m-s37050-r2_GOOG_3833656.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s2k-d256-match14m-s37050-r2_GOOG_3833656.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s2k-d256-match14m-s37050-r2_GOOG_3833656"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-match14m-s37050-r2_GOOG_3833656/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-match14m-s37050-r2_GOOG_3833656/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s2k-d256-match14m-s37050-r2_GOOG_3833656/data_real.tar",
            "selector": null,
            "source_value": 0.122805,
            "report_record_id": "canonical-scores:37712c45-68fd-81df-90ea-e73694dfcb75",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e0-86e9-ff02d54a6b8d",
      "source_row_id": "37712c45-68fd-81e0-86e9-ff02d54a6b8d",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e086e9ff02d54a6b8d",
      "run": "pareto-55m-s12718",
      "stock": null,
      "model_size": "55M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.246744,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.246744",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.190884,
      "reported_ks_context": 0.13977,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:38Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e086e9ff02d54a6b8d",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:38Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:38Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:38Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.246744,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e0-86e9-ff02d54a6b8d",
        "class": "G20",
        "model_evaluation": "pareto-55m-s12718",
        "wws21_mean": 0.246744,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:38Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s12718_GOOG_3253388/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s12718_GOOG_3253388/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:38Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e086e9ff02d54a6b8d",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 951,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916/12718",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s12718_GOOG_3253388.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916 (step 12718)",
            "checkpoint_step": 12718,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s12718_GOOG_3253388.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s12718_GOOG_3253388\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 12718         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.183844+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s12718_GOOG_3253388.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s12718_GOOG_3253388\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 12718         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s12718_GOOG_3253388.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s12718_GOOG_3253388.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s12718_GOOG_3253388"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s12718_GOOG_3253388/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s12718_GOOG_3253388/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s12718_GOOG_3253388/data_real.tar",
            "selector": null,
            "source_value": 0.246744,
            "report_record_id": "canonical-scores:37712c45-68fd-81e0-86e9-ff02d54a6b8d",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e0-8be3-fa4299f65867",
      "source_row_id": "37712c45-68fd-81e0-8be3-fa4299f65867",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e08be3fa4299f65867",
      "run": "j2503382_step22237_24tok",
      "stock": null,
      "model_size": null,
      "run_family": "j2503382",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.184066,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.184066",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.143285,
      "reported_ks_context": 0.103735,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T18:24:47Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2231",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T18:24:47Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T18:24:47Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T18:24:47Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.184066,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e0-8be3-fa4299f65867",
        "class": "G20",
        "model_evaluation": "j2503382_step22237_24tok",
        "wws21_mean": 0.184066,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T18:24:47Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2503382_step22237_24tok/scores/scores_uncond_GOOG_j2503382_step22237_24tok_shard9_integrated_2541176.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2503382_step22237_24tok_GOOG_2541176/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T18:24:47Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2503382_step22237_24tok_GOOG_2541176/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e08be3fa4299f65867",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 952,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "22237",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2503382_7ozaawgb_2503382/22237",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2503382_7ozaawgb_2503382",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2503382_step22237_24tok_GOOG_2541176/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2503382_7ozaawgb_2503382/22237/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2503382_7ozaawgb_2503382/22237",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/LOBS5/checkpoints/j2503382_7ozaawgb_2503382/22237",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "22237",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/LOBS5/checkpoints/j2503382_7ozaawgb_2503382/22237",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2503382_step22237_24tok_GOOG_2541176.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2503382_7ozaawgb_2503382\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2503382_step22237_24tok_GOOG_2541176\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 22237         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2503382_step22237_24tok_GOOG_2541176.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2503382_step22237_24tok/scores/scores_uncond_GOOG_j2503382_step22237_24tok_shard9_integrated_2541176.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2503382_step22237_24tok/scores/scores_uncond_GOOG_j2503382_step22237_24tok_integrated_2541176.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1840658591527453,
        "feature_points": {
          "spread": 0.4002106137516844,
          "ask_cancellation_depth": 0.3040023757858191,
          "bid_cancellation_depth": 0.32590386411647965,
          "limit_ask_order_levels": 0.08811528036392641,
          "limit_bid_order_levels": 0.10315228667013521,
          "ask_cancellation_levels": 0.17407398897656357,
          "bid_cancellation_levels": 0.16068192425067368,
          "vol_per_min": 0.12555309194979558,
          "ofi": 0.08928206809476126,
          "ofi_up": 0.10327803777387094,
          "ofi_stay": 0.08939784564265668,
          "orderbook_imbalance": 0.09386208899414988,
          "ofi_down": 0.09773389866711807,
          "log_inter_arrival_time": 0.02328445502069408,
          "log_time_to_cancel": 0.4155354436791235,
          "ask_volume_touch": 0.1558724205303607,
          "bid_volume_touch": 0.14100840864955536,
          "ask_volume": 0.1363863265076499,
          "bid_volume": 0.19408805819100255,
          "limit_ask_order_depth": 0.3055782622389817,
          "limit_bid_order_depth": 0.33838230235264893
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2503382_step22237_24tok/scores/scores_uncond_GOOG_j2503382_step22237_24tok_integrated_2541176.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.184066,
            "report_record_id": "canonical-scores:37712c45-68fd-81e0-8be3-fa4299f65867",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e0-9b5d-cdf972201982",
      "source_row_id": "37712c45-68fd-81e0-9b5d-cdf972201982",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e09b5dcdf972201982",
      "run": "ctx4k-cond5000",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.167029,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.167029",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.125654,
      "reported_ks_context": 0.091569,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:42Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e09b5dcdf972201982",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:42Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:42Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:42Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.167029,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e0-9b5d-cdf972201982",
        "class": "G20",
        "model_evaluation": "ctx4k-cond5000",
        "wws21_mean": 0.167029,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:42Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond5000/scores/scores_uncond_GOOG_ctx4k-cond5000_integrated_2828768.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond5000_GOOG_2828768/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:42Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond5000_GOOG_2828768/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e09b5dcdf972201982",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 953,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond5000_GOOG_2828768/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 5000,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond5000_GOOG_2828768.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond5000_GOOG_2828768\"         --n_cond_msgs 5000         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond5000_GOOG_2828768.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond5000/scores/scores_uncond_GOOG_ctx4k-cond5000_integrated_2828768.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond5000/scores/scores_uncond_GOOG_ctx4k-cond5000_integrated_2828768.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.16702861173690745,
        "feature_points": {
          "limit_ask_order_levels": 0.06830982863820398,
          "ask_cancellation_levels": 0.05285065598867605,
          "limit_bid_order_levels": 0.015000546360843737,
          "bid_cancellation_levels": 0.02722121004010514,
          "ofi_down": 0.08495340640499881,
          "ofi_up": 0.056763377768314,
          "vol_per_min": 0.08169343221422434,
          "ask_cancellation_depth": 0.34740919549744215,
          "spread": 0.28310167486643095,
          "limit_ask_order_depth": 0.34213787681173297,
          "bid_cancellation_depth": 0.3475529683329825,
          "limit_bid_order_depth": 0.3389519102085089,
          "ofi_stay": 0.046527458282074545,
          "log_time_to_cancel": 0.36535644670195055,
          "ask_volume": 0.3235067587737363,
          "ask_volume_touch": 0.1248053943472898,
          "bid_volume_touch": 0.12276465353428674,
          "orderbook_imbalance": 0.11751746730581902,
          "ofi": 0.04512232264438465,
          "bid_volume": 0.27262614657182327,
          "log_inter_arrival_time": 0.04342811518122798
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond5000/scores/scores_uncond_GOOG_ctx4k-cond5000_integrated_2828768.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.167029,
            "report_record_id": "canonical-scores:37712c45-68fd-81e0-9b5d-cdf972201982",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e0-b799-e0701739c1b6",
      "source_row_id": "37712c45-68fd-81e0-b799-e0701739c1b6",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e0b799e0701739c1b6",
      "run": "j2514440-step135458-L100",
      "stock": null,
      "model_size": null,
      "run_family": "j2514440",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.146141,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.146141",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.117,
      "reported_ks_context": 0.080613,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-01T08:19:43Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2150",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-01T08:19:43Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-01T08:19:43Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-01T08:19:43Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.146141,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e0-b799-e0701739c1b6",
        "class": "G20",
        "model_evaluation": "j2514440-step135458-L100",
        "wws21_mean": 0.146141,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-01T08:19:43Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440-step135458-L100/scores/scores_uncond_GOOG_j2514440-step135458-L100_integrated_2756973.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2514440-step135458-L100_GOOG_2756973/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-01T08:19:43Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2514440-step135458-L100_GOOG_2756973/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e0b799e0701739c1b6",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 954,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2514440-step135458-L100_GOOG_2756973/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=135458 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440-step135458-L100_GOOG_2756973.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2514440-step135458-L100_GOOG_2756973\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 135458         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440-step135458-L100_GOOG_2756973.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440-step135458-L100/scores/scores_uncond_GOOG_j2514440-step135458-L100_integrated_2756973.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440-step135458-L100/scores/scores_uncond_GOOG_j2514440-step135458-L100_integrated_2756973.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.14614115632045693,
        "feature_points": {
          "limit_ask_order_depth": 0.3237034818093223,
          "ask_cancellation_levels": 0.09069416304952893,
          "ofi_down": 0.04720682406592945,
          "limit_bid_order_depth": 0.3396547087832877,
          "limit_ask_order_levels": 0.09213920352446703,
          "bid_cancellation_depth": 0.3190693605349822,
          "vol_per_min": 0.12103699384921424,
          "bid_cancellation_levels": 0.07993591891952845,
          "limit_bid_order_levels": 0.09632042802894862,
          "orderbook_imbalance": 0.11078948501244258,
          "ofi_up": 0.027799061487308936,
          "spread": 0.23441803998943223,
          "ask_cancellation_depth": 0.31148635078558795,
          "bid_volume": 0.12747133996309307,
          "ask_volume_touch": 0.06483376529757598,
          "log_time_to_cancel": 0.3426535276092281,
          "bid_volume_touch": 0.04727015510262201,
          "ofi_stay": 0.05608109177834024,
          "ofi": 0.05502115073174674,
          "ask_volume": 0.1279528077463468,
          "log_inter_arrival_time": 0.053426424660661886
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440-step135458-L100/scores/scores_uncond_GOOG_j2514440-step135458-L100_integrated_2756973.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.146141,
            "report_record_id": "canonical-scores:37712c45-68fd-81e0-b799-e0701739c1b6",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e1-a7bd-e56ce719eee5",
      "source_row_id": "37712c45-68fd-81e1-a7bd-e56ce719eee5",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e1a7bde56ce719eee5",
      "run": "N1b-split-s62183-L100",
      "stock": null,
      "model_size": null,
      "run_family": "N1b",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.214624,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.214624",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.185739,
      "reported_ks_context": 0.137981,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:52Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e1a7bde56ce719eee5",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:52Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:52Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:52Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.214624,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e1-a7bd-e56ce719eee5",
        "class": "G20",
        "model_evaluation": "N1b-split-s62183-L100",
        "wws21_mean": 0.214624,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:52Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100_GOOG_3231569/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100_GOOG_3231569/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:52Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e1a7bde56ce719eee5",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 955,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100_GOOG_3231569.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129 (step 62183)",
            "checkpoint_step": 62183,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100_GOOG_3231569.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1b-split-s62183-L100_GOOG_3231569\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 62183         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263186+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.181399+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100_GOOG_3231569.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1b-split-s62183-L100_GOOG_3231569\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 62183         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100_GOOG_3231569.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100_GOOG_3231569.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1b-split-s62183-L100_GOOG_3231569"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100_GOOG_3231569/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100_GOOG_3231569/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100_GOOG_3231569/data_real.tar",
            "selector": null,
            "source_value": 0.214624,
            "report_record_id": "canonical-scores:37712c45-68fd-81e1-a7bd-e56ce719eee5",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e1-b1a4-e6a4aa145b78",
      "source_row_id": "37712c45-68fd-81e1-b1a4-e6a4aa145b78",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e1b1a4e6a4aa145b78",
      "run": "ticks-gdn-rand",
      "stock": null,
      "model_size": null,
      "run_family": "ticks",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.217249,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.217249",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.170463,
      "reported_ks_context": 0.100704,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:08Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e1b1a4e6a4aa145b78",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:08Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:08Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:08Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.217249,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e1-b1a4-e6a4aa145b78",
        "class": "G20",
        "model_evaluation": "ticks-gdn-rand",
        "wws21_mean": 0.217249,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:08Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-gdn-rand/scores/scores_uncond_GOOG_ticks-gdn-rand_integrated_3278220.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:08Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e1b1a4e6a4aa145b78",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 956,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "0",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137/log_rank0.txt",
            "evidence_line": 753,
            "evidence_text": "[Checkpoint] Loading step=0 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5/0",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5",
            "availability": "not_found",
            "path_scope": "checkpoint_root",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5/0",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. This is the merged soup checkpoint used for inference; source member checkpoints were not independently verified.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-gdn-rand_GOOG_3278220.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-gdn-rand_GOOG_3278220.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-gdn-rand_GOOG_3278220.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-gdn-rand_integrated_3278220.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-rand_GOOG_2888137.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_gdn_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-gdn-rand_GOOG_2888137.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-gdn-rand_GOOG_3278220.out",
            "line": 13,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-gdn-rand_GOOG_3278220.out",
            "line": 17,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/soup-gdn-rand_GOOG_2888137"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ticks-gdn-rand_GOOG_3278220.out",
            "line": 131,
            "text": "  Loading: scores_uncond_GOOG_ticks-gdn-rand_integrated_3278220.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-gdn-rand/scores/scores_uncond_GOOG_ticks-gdn-rand_integrated_3278220.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-gdn-rand/scores/scores_uncond_GOOG_ticks-gdn-rand_integrated_3278220.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.2172494523844653,
        "feature_points": {
          "bid_cancellation_ticks": 0.39551757086795325,
          "ofi_down": 0.09487415202034127,
          "limit_ask_order_depth": 0.3966387634732775,
          "ask_cancellation_depth": 0.380839807503521,
          "ofi_up": 0.05180543631749179,
          "ask_cancellation_ticks": 0.3804307138552474,
          "vol_per_min": 0.06749636730212863,
          "limit_bid_order_depth": 0.39745404832246817,
          "spread": 0.10262407496024911,
          "bid_cancellation_depth": 0.3958948475596038,
          "limit_bid_order_ticks": 0.39740610475257065,
          "limit_ask_order_ticks": 0.39661806939143035,
          "ofi": 0.13396731570720982,
          "log_time_to_cancel": 0.33240261103600766,
          "bid_volume_touch": 0.06623255093652236,
          "bid_volume": 0.09646903561440798,
          "ask_volume_touch": 0.04796538225419064,
          "orderbook_imbalance": 0.08867481964132125,
          "ofi_stay": 0.13565602887630535,
          "ask_volume": 0.07117543342764908,
          "log_inter_arrival_time": 0.1320953662538748
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ticks-gdn-rand/scores/scores_uncond_GOOG_ticks-gdn-rand_integrated_3278220.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.217249,
            "report_record_id": "canonical-scores:37712c45-68fd-81e1-b1a4-e6a4aa145b78",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e2-9ffe-edd051283583",
      "source_row_id": "37712c45-68fd-81e2-9ffe-edd051283583",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e29ffeedd051283583",
      "run": "R1-s4k-d256-raw-s35580",
      "stock": null,
      "model_size": null,
      "run_family": "R1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.150368,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.150368",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.177608,
      "reported_ks_context": 0.097699,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:43:46Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e29ffeedd051283583",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:43:46Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:43:46Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:43:46Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.150368,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e2-9ffe-edd051283583",
        "class": "G20",
        "model_evaluation": "R1-s4k-d256-raw-s35580",
        "wws21_mean": 0.150368,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:43:46Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s35580_GOOG_3745460/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s35580_GOOG_3745460/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:43:46Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e29ffeedd051283583",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 957,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096/35580",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s35580_GOOG_3745460.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096 (step 35580)",
            "checkpoint_step": 35580,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s35580_GOOG_3745460.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s35580_GOOG_3745460\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 35580         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.136864+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s35580_GOOG_3745460.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3739096_qp7lvh54_3739096\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s35580_GOOG_3745460\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 35580         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s35580_GOOG_3745460.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_R1-s4k-d256-raw-s35580_GOOG_3745460.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/R1-s4k-d256-raw-s35580_GOOG_3745460"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s35580_GOOG_3745460/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s35580_GOOG_3745460/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/R1-s4k-d256-raw-s35580_GOOG_3745460/data_real.tar",
            "selector": null,
            "source_value": 0.150368,
            "report_record_id": "canonical-scores:37712c45-68fd-81e2-9ffe-edd051283583",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e2-af73-fa2de7649fa2",
      "source_row_id": "37712c45-68fd-81e2-af73-fa2de7649fa2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e2af73fa2de7649fa2",
      "run": "soup-ctxt4k-rand",
      "stock": null,
      "model_size": null,
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.156259,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.156259",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.122767,
      "reported_ks_context": 0.085794,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:27Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e2af73fa2de7649fa2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:27Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:27Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:27Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.156259,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e2-af73-fa2de7649fa2",
        "class": "G20",
        "model_evaluation": "soup-ctxt4k-rand",
        "wws21_mean": 0.156259,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:27Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:27Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e2af73fa2de7649fa2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 958,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt4k-rand_GOOG_2888144.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt4k-rand_GOOG_2888144.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262982+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134305+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt4k-rand_GOOG_2888144.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt4k-rand_GOOG_2888144.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-ctxt4k-rand_GOOG_2888144.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-ctxt4k-rand_GOOG_2888144/data_real.tar",
            "selector": null,
            "source_value": 0.156259,
            "report_record_id": "canonical-scores:37712c45-68fd-81e2-af73-fa2de7649fa2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e4-8296-d8c119b9c817",
      "source_row_id": "37712c45-68fd-81e4-8296-d8c119b9c817",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e48296d8c119b9c817",
      "run": "soup-120m-rand",
      "stock": null,
      "model_size": "120M",
      "run_family": "soup",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.14279,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.14279",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.11308,
      "reported_ks_context": 0.07565,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:35Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e48296d8c119b9c817",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:35Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:35Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:35Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.14279,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e4-8296-d8c119b9c817",
        "class": "G20",
        "model_evaluation": "soup-120m-rand",
        "wws21_mean": 0.14279,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:35Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:35Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e48296d8c119b9c817",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 959,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-120m-rand_GOOG_2887609.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-120m-rand_GOOG_2887609.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263216+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.185143+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-120m-rand_GOOG_2887609.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_120M_131k_135k\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-120m-rand_GOOG_2887609.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_soup-120m-rand_GOOG_2887609.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/soup-120m-rand_GOOG_2887609/data_real.tar",
            "selector": null,
            "source_value": 0.14279,
            "report_record_id": "canonical-scores:37712c45-68fd-81e4-8296-d8c119b9c817",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e4-879e-ecf07eeb8a06",
      "source_row_id": "37712c45-68fd-81e4-879e-ecf07eeb8a06",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e4879eecf07eeb8a06",
      "run": "G10-l3k-to-l10-arm",
      "stock": null,
      "model_size": null,
      "run_family": "G10",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.464743,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.464743",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.35806,
      "reported_ks_context": 0.307338,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:17Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e4879eecf07eeb8a06",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:17Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:17Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:17Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.464743,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e4-879e-ecf07eeb8a06",
        "class": "G20",
        "model_evaluation": "G10-l3k-to-l10-arm",
        "wws21_mean": 0.464743,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:17Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-to-l10-arm_GOOG_3508934/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-to-l10-arm_GOOG_3508934/data_real.tar;/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-to-l10-arm_GOOG_3512718/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:17Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 2048 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e4879eecf07eeb8a06",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 960,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j3484455_tovfdcv5_3484455/132248",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-to-l10-arm_GOOG_3508934.out",
            "evidence_line": 8,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j3484455_tovfdcv5_3484455 (step 132248)",
            "checkpoint_step": 132248,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j3484455_tovfdcv5_3484455",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-to-l10-arm_GOOG_3508934.out",
                "evidence_line": 22,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j3484455_tovfdcv5_3484455\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l3k-to-l10-arm_GOOG_3508934\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 2048         --batch_size 64         --checkpoint_step 132248         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.185163+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j3484455_tovfdcv5_3484455",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j3484455_tovfdcv5_3484455"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-to-l10-arm_GOOG_3508934.out",
            "line": 22,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j3484455_tovfdcv5_3484455\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l3k-to-l10-arm_GOOG_3508934\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 2048         --batch_size 64         --checkpoint_step 132248         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-to-l10-arm_GOOG_3508934.out",
            "line": 14,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l3k-to-l10-arm_GOOG_3508934.out",
            "line": 19,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l3k-to-l10-arm_GOOG_3508934"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-to-l10-arm_GOOG_3508934/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-to-l10-arm_GOOG_3508934/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l3k-to-l10-arm_GOOG_3508934/data_real.tar",
            "selector": null,
            "source_value": 0.464743,
            "report_record_id": "canonical-scores:37712c45-68fd-81e4-879e-ecf07eeb8a06",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e5-bfbe-cd326a1d6623",
      "source_row_id": "37712c45-68fd-81e5-bfbe-cd326a1d6623",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e5bfbecd326a1d6623",
      "run": "ctx4k-cond2500",
      "stock": null,
      "model_size": null,
      "run_family": "ctx4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.153838,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.153838",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.123223,
      "reported_ks_context": 0.086852,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:55Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e5bfbecd326a1d6623",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:55Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:55Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:55Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.153838,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e5-bfbe-cd326a1d6623",
        "class": "G20",
        "model_evaluation": "ctx4k-cond2500",
        "wws21_mean": 0.153838,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:55Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond2500/scores/scores_uncond_GOOG_ctx4k-cond2500_integrated_2828763.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond2500_GOOG_2828763/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:55Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond2500_GOOG_2828763/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e5bfbecd326a1d6623",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 961,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "102965",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/ctx4k-cond2500_GOOG_2828763/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[Checkpoint] Loading step=102965 from /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167 (partial_restore=True)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/102965",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 2500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond2500_GOOG_2828763.out",
            "line": 20,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py                  --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/ctx4k-cond2500_GOOG_2828763\"         --n_cond_msgs 2500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 102965         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_ctx4k-cond2500_GOOG_2828763.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond2500/scores/scores_uncond_GOOG_ctx4k-cond2500_integrated_2828763.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond2500/scores/scores_uncond_GOOG_ctx4k-cond2500_integrated_2828763.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.15383832764484484,
        "feature_points": {
          "bid_cancellation_depth": 0.31623520772280983,
          "vol_per_min": 0.05664886218561999,
          "ofi_down": 0.07462868317275047,
          "spread": 0.2691851738886019,
          "orderbook_imbalance": 0.0978258937422575,
          "limit_ask_order_levels": 0.06531661728853444,
          "ask_volume_touch": 0.09800175026559543,
          "bid_cancellation_levels": 0.020946710443268225,
          "ask_volume": 0.27779373562367426,
          "ask_cancellation_levels": 0.06796561084365198,
          "limit_bid_order_depth": 0.33921564602008514,
          "limit_bid_order_levels": 0.018798428246152483,
          "ofi_up": 0.0703900256539477,
          "bid_volume": 0.31764154635556596,
          "ofi_stay": 0.056412913616186576,
          "limit_ask_order_depth": 0.3110831319988321,
          "ask_cancellation_depth": 0.21449822029315996,
          "ofi": 0.05513023663322603,
          "log_time_to_cancel": 0.36612616354982563,
          "bid_volume_touch": 0.08832654375178497,
          "log_inter_arrival_time": 0.048433779246211124
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_ctx4k-cond2500/scores/scores_uncond_GOOG_ctx4k-cond2500_integrated_2828763.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.153838,
            "report_record_id": "canonical-scores:37712c45-68fd-81e5-bfbe-cd326a1d6623",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e6-a7a2-dcd67a7f902e",
      "source_row_id": "37712c45-68fd-81e6-a7a2-dcd67a7f902e",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e6a7a2dcd67a7f902e",
      "run": "46m-s2k-d256-nobook-hier-c1750g250",
      "stock": null,
      "model_size": "46M",
      "run_family": "46m",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.115489,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.115489",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.162222,
      "reported_ks_context": 0.102975,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:03Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e6a7a2dcd67a7f902e",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:03Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:03Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:03Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.115489,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e6-a7a2-dcd67a7f902e",
        "class": "G20",
        "model_evaluation": "46m-s2k-d256-nobook-hier-c1750g250",
        "wws21_mean": 0.115489,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:03Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:03Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e6a7a2dcd67a7f902e",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 962,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301/37610",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301 (step 37610)",
            "checkpoint_step": 37610,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048\"         --n_cond_msgs 1750         --n_gen_msgs 250         --n_sequences 1024         --batch_size 16         --checkpoint_step 37610         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.174479+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 1750,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3865301_xb1bpgz0_3865301\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048\"         --n_cond_msgs 1750         --n_gen_msgs 250         --n_sequences 1024         --batch_size 16         --checkpoint_step 37610         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/46m-s2k-d256-nobook-hier-c1750g250_GOOG_3876048/data_real.tar",
            "selector": null,
            "source_value": 0.115489,
            "report_record_id": "canonical-scores:37712c45-68fd-81e6-a7a2-dcd67a7f902e",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e7-9582-efa2fe857d97",
      "source_row_id": "37712c45-68fd-81e7-9582-efa2fe857d97",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e79582efa2fe857d97",
      "run": "adamw-32n-s29706",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.139635,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.139635",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.11061,
      "reported_ks_context": 0.074885,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:18Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e79582efa2fe857d97",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:18Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:18Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:18Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.139635,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e7-9582-efa2fe857d97",
        "class": "G20",
        "model_evaluation": "adamw-32n-s29706",
        "wws21_mean": 0.139635,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:18Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29706_GOOG_2871389/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29706_GOOG_2871389/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:18Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e79582efa2fe857d97",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 963,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29706_GOOG_2871389.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 29706)",
            "checkpoint_step": 29706,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29706_GOOG_2871389.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s29706_GOOG_2871389\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 29706         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29706_GOOG_2871389.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s29706_GOOG_2871389\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 29706         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29706_GOOG_2871389.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s29706_GOOG_2871389.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s29706_GOOG_2871389"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29706_GOOG_2871389/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29706_GOOG_2871389/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s29706_GOOG_2871389/data_real.tar",
            "selector": null,
            "source_value": 0.139635,
            "report_record_id": "canonical-scores:37712c45-68fd-81e7-9582-efa2fe857d97",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81e8-9372-ee0f35689039",
      "source_row_id": "37712c45-68fd-81e8-9372-ee0f35689039",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81e89372ee0f35689039",
      "run": "I3_120M_step12815",
      "stock": null,
      "model_size": "120M",
      "run_family": "I3",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.245776,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.245776",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.194522,
      "reported_ks_context": 0.126315,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:24Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81e89372ee0f35689039",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:24Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:24Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:24Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.245776,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81e8-9372-ee0f35689039",
        "class": "G20",
        "model_evaluation": "I3_120M_step12815",
        "wws21_mean": 0.245776,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:24Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3_120M_step12815/scores/scores_uncond_GOOG_I3_120M_step12815_shard5_integrated_2566038.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/I3_120M_step12815_GOOG_2566038/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:24Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/I3_120M_step12815_GOOG_2566038/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81e89372ee0f35689039",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 964,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "12815",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/12815",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/I3_120M_step12815_GOOG_2566038/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/12815/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/12815",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/12815",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/12815",
            "availability": "not_found",
            "path_scope": "checkpoint_step",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/12815",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            },
            "relocation_candidates": [
              {
                "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_I3_effective_bsz_DELETING_91047/checkpoints/j2518919_zqmh37u1_2518919/12815",
                "original_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919/12815",
                "availability": "exists",
                "entries": [
                  "state",
                  "_CHECKPOINT_METADATA",
                  "metadata"
                ],
                "type": "directory",
                "path_kind": "directory",
                "note": "Possible relocated copy under _DELETING_91047; exact run and step directory exists, identity unconfirmed.",
                "identity_status": "unconfirmed_identity",
                "caveat": "Observed _DELETING_91047 directory. Exact run and step names match, but no authoritative rename record was inspected."
              }
            ]
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately. An existing exact run/step candidate was found under an observed _DELETING_91047 directory. It remains a candidate because no authoritative rename record was inspected."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_I3_120M_step12815_GOOG_2566038.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_I3_effective_bsz/checkpoints/j2518919_zqmh37u1_2518919\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/I3_120M_step12815_GOOG_2566038\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 12815         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_I3_120M_step12815_GOOG_2566038.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3_120M_step12815/scores/scores_uncond_GOOG_I3_120M_step12815_shard5_integrated_2566038.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3_120M_step12815/scores/scores_uncond_GOOG_I3_120M_step12815_integrated_2566038.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.24577585397171353,
        "feature_points": {
          "spread": 0.6262102880814937,
          "ask_cancellation_depth": 0.40303938304574516,
          "bid_cancellation_depth": 0.35106798472862133,
          "limit_ask_order_levels": 0.09873947827423497,
          "limit_bid_order_levels": 0.10971587856361037,
          "ask_cancellation_levels": 0.15754304643406772,
          "bid_cancellation_levels": 0.15010075328795877,
          "vol_per_min": 0.14485490642825485,
          "ofi": 0.13536848548596203,
          "ofi_up": 0.23687195030478791,
          "ofi_stay": 0.1356145174142119,
          "orderbook_imbalance": 0.10335313446989874,
          "ofi_down": 0.22487170761543668,
          "log_inter_arrival_time": 0.05862108518919707,
          "log_time_to_cancel": 0.5248352895501427,
          "ask_volume_touch": 0.21528368515280544,
          "bid_volume_touch": 0.23387391487252,
          "ask_volume": 0.21628700721078753,
          "bid_volume": 0.2953338657374064,
          "limit_ask_order_depth": 0.40717125563373585,
          "limit_bid_order_depth": 0.3325353159251046
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_I3_120M_step12815/scores/scores_uncond_GOOG_I3_120M_step12815_integrated_2566038.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.245776,
            "report_record_id": "canonical-scores:37712c45-68fd-81e8-9372-ee0f35689039",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ea-a33b-fd1654cbb690",
      "source_row_id": "37712c45-68fd-81ea-a33b-fd1654cbb690",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81eaa33bfd1654cbb690",
      "run": "phaseA-r025-a-s34250-c250g250",
      "stock": null,
      "model_size": null,
      "run_family": "phaseA",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.127919,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.127919",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.1829,
      "reported_ks_context": 0.109148,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:11Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81eaa33bfd1654cbb690",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:11Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:11Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:11Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.127919,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ea-a33b-fd1654cbb690",
        "class": "G20",
        "model_evaluation": "phaseA-r025-a-s34250-c250g250",
        "wws21_mean": 0.127919,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:11Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c250g250_GOOG_3701968/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c250g250_GOOG_3701968/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:11Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81eaa33bfd1654cbb690",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 965,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346/34250",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c250g250_GOOG_3701968.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346 (step 34250)",
            "checkpoint_step": 34250,
            "logged_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c250g250_GOOG_3701968.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r025-a-s34250-c250g250_GOOG_3701968\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 34250         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 25,
                "evidence_text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.160721+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c250g250_GOOG_3701968.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3696346_rixarh6v_3696346\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseA-r025-a-s34250-c250g250_GOOG_3701968\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 34250         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c250g250_GOOG_3701968.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseA-r025-a-s34250-c250g250_GOOG_3701968.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseA-r025-a-s34250-c250g250_GOOG_3701968"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c250g250_GOOG_3701968/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c250g250_GOOG_3701968/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseA-r025-a-s34250-c250g250_GOOG_3701968/data_real.tar",
            "selector": null,
            "source_value": 0.127919,
            "report_record_id": "canonical-scores:37712c45-68fd-81ea-a33b-fd1654cbb690",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ea-b0f9-f6fea5a2063c",
      "source_row_id": "37712c45-68fd-81ea-b0f9-f6fea5a2063c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81eab0f9f6fea5a2063c",
      "run": "s4k-d512-32n-step23460",
      "stock": null,
      "model_size": null,
      "run_family": "s4k",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.108851,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.108851",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.166124,
      "reported_ks_context": 0.10113,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:01Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81eab0f9f6fea5a2063c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:01Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:01Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:01Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.108851,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ea-b0f9-f6fea5a2063c",
        "class": "G20",
        "model_evaluation": "s4k-d512-32n-step23460",
        "wws21_mean": 0.108851,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:01Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step23460_GOOG_3763547/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step23460_GOOG_3763547/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:01Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81eab0f9f6fea5a2063c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 966,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036/23460",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step23460_GOOG_3763547.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036 (step 23460)",
            "checkpoint_step": 23460,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step23460_GOOG_3763547.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step23460_GOOG_3763547\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 23460         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.176082+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step23460_GOOG_3763547.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3760036_g4py8axs_3760036\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step23460_GOOG_3763547\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 23460         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step23460_GOOG_3763547.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_s4k-d512-32n-step23460_GOOG_3763547.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/s4k-d512-32n-step23460_GOOG_3763547"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step23460_GOOG_3763547/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step23460_GOOG_3763547/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/s4k-d512-32n-step23460_GOOG_3763547/data_real.tar",
            "selector": null,
            "source_value": 0.108851,
            "report_record_id": "canonical-scores:37712c45-68fd-81ea-b0f9-f6fea5a2063c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ec-8bd7-e9857defb156",
      "source_row_id": "37712c45-68fd-81ec-8bd7-e9857defb156",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ec8bd7e9857defb156",
      "run": "j2496000_360M_step10990",
      "stock": null,
      "model_size": "360M",
      "run_family": "j2496000",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.176756,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.176756",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.140613,
      "reported_ks_context": 0.100679,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T04:07:35Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2276",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T04:07:35Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T04:07:35Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T04:07:35Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.176756,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ec-8bd7-e9857defb156",
        "class": "G20",
        "model_evaluation": "j2496000_360M_step10990",
        "wws21_mean": 0.176756,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T04:07:35Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496000_360M_step10990/scores/scores_uncond_GOOG_j2496000_360M_step10990_shard17_integrated_2552671.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496000_360M_step10990_GOOG_2552671/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T04:07:35Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496000_360M_step10990_GOOG_2552671/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ec8bd7e9857defb156",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 967,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "10990",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2496000_e6i8kq38_2496000/10990",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2496000_e6i8kq38_2496000",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496000_360M_step10990_GOOG_2552671/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2496000_e6i8kq38_2496000/10990/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2496000_e6i8kq38_2496000/10990",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/LOBS5/checkpoints/j2496000_e6i8kq38_2496000/10990",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "10066",
                "10225",
                "10372",
                "10522",
                "10669",
                "10831",
                "10990",
                "9440",
                "9657",
                "9864",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/LOBS5/checkpoints/j2496000_e6i8kq38_2496000/10990",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496000_360M_step10990_GOOG_2552671.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2496000_e6i8kq38_2496000\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2496000_360M_step10990_GOOG_2552671\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 10990         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496000_360M_step10990_GOOG_2552671.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496000_360M_step10990/scores/scores_uncond_GOOG_j2496000_360M_step10990_shard17_integrated_2552671.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496000_360M_step10990/scores/scores_uncond_GOOG_j2496000_360M_step10990_integrated_2552671.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.17675551366163814,
        "feature_points": {
          "spread": 0.3887886620518117,
          "ask_cancellation_depth": 0.30214112947044036,
          "bid_cancellation_depth": 0.3205815102819281,
          "limit_ask_order_levels": 0.08452160721176062,
          "limit_bid_order_levels": 0.09600288654738116,
          "ask_cancellation_levels": 0.1516549647655809,
          "bid_cancellation_levels": 0.14695324812025565,
          "vol_per_min": 0.13860411820313714,
          "ofi": 0.06600560104888364,
          "ofi_up": 0.11460469441991443,
          "ofi_stay": 0.06594238597243984,
          "orderbook_imbalance": 0.1105272615230468,
          "ofi_down": 0.0831039643386743,
          "log_inter_arrival_time": 0.015203642486317283,
          "log_time_to_cancel": 0.45206321160447716,
          "ask_volume_touch": 0.1170397927524142,
          "bid_volume_touch": 0.14046053253593863,
          "ask_volume": 0.10844126295102231,
          "bid_volume": 0.15996321354805404,
          "limit_ask_order_depth": 0.3159070608948041,
          "limit_bid_order_depth": 0.3333550361661184
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496000_360M_step10990/scores/scores_uncond_GOOG_j2496000_360M_step10990_integrated_2552671.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.176756,
            "report_record_id": "canonical-scores:37712c45-68fd-81ec-8bd7-e9857defb156",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ee-894a-f1ad6b270f0a",
      "source_row_id": "37712c45-68fd-81ee-894a-f1ad6b270f0a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ee894af1ad6b270f0a",
      "run": "phaseB-m3-14m-s500-d64-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.136377,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.136377",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.181776,
      "reported_ks_context": 0.104741,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81ee894af1ad6b270f0a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.136377,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ee-894a-f1ad6b270f0a",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s500-d64-b-avg-c250g250",
        "wws21_mean": 0.136377,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ee894af1ad6b270f0a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 968,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951/log_rank0.txt",
            "evidence_line": 586,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703790_mysq7uya_3703790",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703790_mysq7uya_3703790\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703790_mysq7uya_3703790 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951/log_rank0.txt",
                "evidence_line": 588,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790",
            "availability_checked_at": "2026-09-12T06:13:15.364479+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703790_mysq7uya_3703790/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703790_mysq7uya_3703790\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d64-b-avg-c250g250_GOOG_3713951/data_real.tar",
            "selector": null,
            "source_value": 0.136377,
            "report_record_id": "canonical-scores:37712c45-68fd-81ee-894a-f1ad6b270f0a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ee-8c22-c3b987cb3f32",
      "source_row_id": "37712c45-68fd-81ee-8c22-c3b987cb3f32",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ee8c22c3b987cb3f32",
      "run": "pareto-55m-s2037",
      "stock": null,
      "model_size": "55M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.338971,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.338971",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.275695,
      "reported_ks_context": 0.18845,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:40:03Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81ee8c22c3b987cb3f32",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:40:03Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:40:03Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:40:03Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.338971,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ee-8c22-c3b987cb3f32",
        "class": "G20",
        "model_evaluation": "pareto-55m-s2037",
        "wws21_mean": 0.338971,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:40:03Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s2037_GOOG_3253386/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s2037_GOOG_3253386/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:40:03Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ee8c22c3b987cb3f32",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 969,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916/2037",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s2037_GOOG_3253386.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916 (step 2037)",
            "checkpoint_step": 2037,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s2037_GOOG_3253386.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s2037_GOOG_3253386\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 2037         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.183844+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s2037_GOOG_3253386.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s2037_GOOG_3253386\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 2037         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s2037_GOOG_3253386.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s2037_GOOG_3253386.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s2037_GOOG_3253386"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s2037_GOOG_3253386/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s2037_GOOG_3253386/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s2037_GOOG_3253386/data_real.tar",
            "selector": null,
            "source_value": 0.338971,
            "report_record_id": "canonical-scores:37712c45-68fd-81ee-8c22-c3b987cb3f32",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ee-9a75-fe9c0ab9fe3a",
      "source_row_id": "37712c45-68fd-81ee-9a75-fe9c0ab9fe3a",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ee9a75fe9c0ab9fe3a",
      "run": "H2-55M-ctx4k-cond5500",
      "stock": null,
      "model_size": "55M",
      "run_family": "H2",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.127063,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.127063",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.111322,
      "reported_ks_context": 0.073743,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:34:21Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81ee9a75fe9c0ab9fe3a",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:34:21Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:34:21Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:34:21Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.127063,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ee-9a75-fe9c0ab9fe3a",
        "class": "G20",
        "model_evaluation": "H2-55M-ctx4k-cond5500",
        "wws21_mean": 0.127063,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:34:21Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond5500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond5500_shard1_integrated_2632449.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond5500_GOOG_2632449/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:34:21Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond5500_GOOG_2632449/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ee9a75fe9c0ab9fe3a",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 970,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "100378",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond5500_GOOG_2632449/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "100378",
                "102965",
                "17650",
                "20235",
                "22820",
                "35750",
                "38335",
                "79688",
                "82272",
                "84859",
                "87443",
                "90029",
                "92616",
                "95204",
                "97793",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167/100378",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 5500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond5500_GOOG_2632449.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H2-context-scale/checkpoints/j2504167_y0c4j6l3_2504167\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H2-55M-ctx4k-cond5500_GOOG_2632449\"         --n_cond_msgs 5500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 6         --checkpoint_step 100378         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H2-55M-ctx4k-cond5500_GOOG_2632449.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond5500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond5500_shard1_integrated_2632449.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond5500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond5500_integrated_2632449.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.12706281467764885,
        "feature_points": {
          "spread": 0.13612815829535985,
          "ask_cancellation_depth": 0.2439607407934204,
          "bid_cancellation_depth": 0.3206057635047955,
          "limit_ask_order_levels": 0.01756772280341462,
          "limit_bid_order_levels": 0.028914168458378468,
          "ask_cancellation_levels": 0.02970963378247024,
          "bid_cancellation_levels": 0.05046410454836208,
          "vol_per_min": 0.09192903434013379,
          "ofi": 0.04322675327351539,
          "ofi_up": 0.06533208272034188,
          "ofi_stay": 0.04385811750392487,
          "orderbook_imbalance": 0.11257394078185758,
          "ofi_down": 0.04427606272351616,
          "log_inter_arrival_time": 0.026509096602340677,
          "log_time_to_cancel": 0.3586683755784429,
          "ask_volume_touch": 0.11701812251669572,
          "bid_volume_touch": 0.09922849753746889,
          "ask_volume": 0.10675777792598745,
          "bid_volume": 0.14267285421266504,
          "limit_ask_order_depth": 0.2633847440119624,
          "limit_bid_order_depth": 0.32553335631557206
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H2-55M-ctx4k-cond5500/scores/scores_uncond_GOOG_H2-55M-ctx4k-cond5500_integrated_2632449.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.127063,
            "report_record_id": "canonical-scores:37712c45-68fd-81ee-9a75-fe9c0ab9fe3a",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ee-bf90-ee4f4d7fb866",
      "source_row_id": "37712c45-68fd-81ee-bf90-ee4f4d7fb866",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81eebf90ee4f4d7fb866",
      "run": "phaseB-m3-14m-s1000-d64-b-avg-c250g250",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.179586,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.179586",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.18479,
      "reported_ks_context": 0.100531,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:51Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81eebf90ee4f4d7fb866",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:51Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:51Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:51Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.179586,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ee-bf90-ee4f4d7fb866",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s1000-d64-b-avg-c250g250",
        "wws21_mean": 0.179586,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:51Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:51Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81eebf90ee4f4d7fb866",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 971,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933/log_rank0.txt",
            "evidence_line": 580,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703984_lxjyvwfw_3703984",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703984_lxjyvwfw_3703984\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703984_lxjyvwfw_3703984 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933/log_rank0.txt",
                "evidence_line": 582,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984",
            "availability_checked_at": "2026-09-12T06:13:15.372948+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703984_lxjyvwfw_3703984/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703984_lxjyvwfw_3703984\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s1000-d64-b-avg-c250g250_GOOG_3713933/data_real.tar",
            "selector": null,
            "source_value": 0.179586,
            "report_record_id": "canonical-scores:37712c45-68fd-81ee-bf90-ee4f4d7fb866",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81ef-9b06-eb74e46ab093",
      "source_row_id": "37712c45-68fd-81ef-9b06-eb74e46ab093",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81ef9b06eb74e46ab093",
      "run": "genlen-ctx4k-c500-g3000",
      "stock": null,
      "model_size": null,
      "run_family": "genlen",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.294359,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.294359",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.227671,
      "reported_ks_context": 0.169205,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:54Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81ef9b06eb74e46ab093",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:54Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:54Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:54Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.294359,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81ef-9b06-eb74e46ab093",
        "class": "G20",
        "model_evaluation": "genlen-ctx4k-c500-g3000",
        "wws21_mean": 0.294359,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:54Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g3000_GOOG_3405788/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g3000_GOOG_3405788/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:35:54Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81ef9b06eb74e46ab093",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 972,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g3000_GOOG_3405788.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5 (step 0)",
            "checkpoint_step": 0,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g3000_GOOG_3405788.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g3000_GOOG_3405788\"         --n_cond_msgs 500         --n_gen_msgs 3000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.262982+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.134305+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented. This is the merged soup checkpoint used for inference; source component checkpoints were not enumerated in these logs."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 3000,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g3000_GOOG_3405788.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K3_Mamba2         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/soup_ctxt4k_last5\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g3000_GOOG_3405788\"         --n_cond_msgs 500         --n_gen_msgs 3000         --n_sequences 1024         --batch_size 64         --checkpoint_step 0         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g3000_GOOG_3405788.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_genlen-ctx4k-c500-g3000_GOOG_3405788.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/genlen-ctx4k-c500-g3000_GOOG_3405788"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g3000_GOOG_3405788/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g3000_GOOG_3405788/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/genlen-ctx4k-c500-g3000_GOOG_3405788/data_real.tar",
            "selector": null,
            "source_value": 0.294359,
            "report_record_id": "canonical-scores:37712c45-68fd-81ef-9b06-eb74e46ab093",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f0-9669-d85e96cc1c99",
      "source_row_id": "37712c45-68fd-81f0-9669-d85e96cc1c99",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f09669d85e96cc1c99",
      "run": "J1-anchor1-61k",
      "stock": null,
      "model_size": null,
      "run_family": "J1",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.181003,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.181003",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.142444,
      "reported_ks_context": 0.10169,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:47Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81f09669d85e96cc1c99",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:47Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:47Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:47Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.181003,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f0-9669-d85e96cc1c99",
        "class": "G20",
        "model_evaluation": "J1-anchor1-61k",
        "wws21_mean": 0.181003,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:47Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor1-61k/scores/scores_uncond_GOOG_J1-anchor1-61k_shard0_integrated_2654619.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor1-61k_GOOG_2654619/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:47Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor1-61k_GOOG_2654619/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f09669d85e96cc1c99",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 973,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "61037",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975/61037",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/J1-anchor1-61k_GOOG_2654619/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975/61037/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975/61037",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975/61037",
              "availability": "not_found",
              "entries": [],
              "root_entries": []
            },
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975/61037",
            "availability": "not_found",
            "path_scope": "checkpoint_step",
            "resolution": "historical_path_retained",
            "current_candidate": {
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975/61037",
              "availability": "not_found",
              "resolution": "recorded_prefix_mapping_candidate"
            }
          }
        ],
        "note": "The exact historical checkpoint is identified, but that path is now missing. Current mapped candidate status is recorded separately."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_J1-anchor1-61k_GOOG_2654619.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J1-sparse-book-anchoring/checkpoints/j2633975_gao5ok51_2633975\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/J1-anchor1-61k_GOOG_2654619\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 61037         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_J1-anchor1-61k_GOOG_2654619.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor1-61k/scores/scores_uncond_GOOG_J1-anchor1-61k_shard0_integrated_2654619.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor1-61k/scores/scores_uncond_GOOG_J1-anchor1-61k_integrated_2654619.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.18100337157777285,
        "feature_points": {
          "spread": 0.41280444434377184,
          "ask_cancellation_depth": 0.2924692790826185,
          "bid_cancellation_depth": 0.3584449807050637,
          "limit_ask_order_levels": 0.06951280344854491,
          "limit_bid_order_levels": 0.04026775478784918,
          "ask_cancellation_levels": 0.06298233244771743,
          "bid_cancellation_levels": 0.04535464650617426,
          "vol_per_min": 0.19014947074061572,
          "ofi": 0.07605211912098177,
          "ofi_up": 0.13045932308491232,
          "ofi_stay": 0.07714381498086037,
          "orderbook_imbalance": 0.10349152914360994,
          "ofi_down": 0.04170024489270416,
          "log_inter_arrival_time": 0.07057459373965509,
          "log_time_to_cancel": 0.5690156673479282,
          "ask_volume_touch": 0.16274161245235413,
          "bid_volume_touch": 0.10754159675950205,
          "ask_volume": 0.17154762311094773,
          "bid_volume": 0.14800529810859137,
          "limit_ask_order_depth": 0.29527044302404826,
          "limit_bid_order_depth": 0.375541225304779
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_J1-anchor1-61k/scores/scores_uncond_GOOG_J1-anchor1-61k_integrated_2654619.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.181003,
            "report_record_id": "canonical-scores:37712c45-68fd-81f0-9669-d85e96cc1c99",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f0-a867-fb11d0c73b68",
      "source_row_id": "37712c45-68fd-81f0-a867-fb11d0c73b68",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f0a867fb11d0c73b68",
      "run": "phaseB-m3-14m-s500-d256-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.161571,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.161571",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.201948,
      "reported_ks_context": 0.114368,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:38:12Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81f0a867fb11d0c73b68",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:38:12Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:38:12Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:38:12Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.161571,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f0-a867-fb11d0c73b68",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s500-d256-b-avg-c500g500",
        "wws21_mean": 0.161571,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:38:12Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:38:12Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f0a867fb11d0c73b68",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 974,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950/log_rank0.txt",
            "evidence_line": 656,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3704411_2qztmsnp_3704411",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704411_2qztmsnp_3704411\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3704411_2qztmsnp_3704411 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950/log_rank0.txt",
                "evidence_line": 658,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411",
            "availability_checked_at": "2026-09-12T06:13:15.374336+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3704411_2qztmsnp_3704411/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3704411_2qztmsnp_3704411\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s500-d256-b-avg-c500g500_GOOG_3713950/data_real.tar",
            "selector": null,
            "source_value": 0.161571,
            "report_record_id": "canonical-scores:37712c45-68fd-81f0-a867-fb11d0c73b68",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f0-b47e-caf5814bfa16",
      "source_row_id": "37712c45-68fd-81f0-b47e-caf5814bfa16",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f0b47ecaf5814bfa16",
      "run": "pareto-360m-adamw32n-s16970",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.168935,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.168935",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.131408,
      "reported_ks_context": 0.092556,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:29Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81f0b47ecaf5814bfa16",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:29Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:29Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:29Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.168935,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f0-b47e-caf5814bfa16",
        "class": "G20",
        "model_evaluation": "pareto-360m-adamw32n-s16970",
        "wws21_mean": 0.168935,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:29Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s16970_GOOG_3253405/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s16970_GOOG_3253405/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:29Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f0b47ecaf5814bfa16",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 975,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s16970_GOOG_3253405.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 16970)",
            "checkpoint_step": 16970,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s16970_GOOG_3253405.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw32n-s16970_GOOG_3253405\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 16970         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s16970_GOOG_3253405.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw32n-s16970_GOOG_3253405\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 16970         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s16970_GOOG_3253405.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw32n-s16970_GOOG_3253405.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw32n-s16970_GOOG_3253405"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s16970_GOOG_3253405/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s16970_GOOG_3253405/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw32n-s16970_GOOG_3253405/data_real.tar",
            "selector": null,
            "source_value": 0.168935,
            "report_record_id": "canonical-scores:37712c45-68fd-81f0-b47e-caf5814bfa16",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f1-82d7-ffd238f85e0f",
      "source_row_id": "37712c45-68fd-81f1-82d7-ffd238f85e0f",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f182d7ffd238f85e0f",
      "run": "j2504227_step22872_24tok",
      "stock": null,
      "model_size": null,
      "run_family": "j2504227",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.175188,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.175188",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.140773,
      "reported_ks_context": 0.101192,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T19:42:09Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2220",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T19:42:09Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T19:42:09Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T19:42:09Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.175188,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f1-82d7-ffd238f85e0f",
        "class": "G20",
        "model_evaluation": "j2504227_step22872_24tok",
        "wws21_mean": 0.175188,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T19:42:09Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504227_step22872_24tok/scores/scores_uncond_GOOG_j2504227_step22872_24tok_integrated_2512979.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504227_step22872_24tok_GOOG_2512979/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T19:42:09Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504227_step22872_24tok_GOOG_2512979/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f182d7ffd238f85e0f",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 976,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "22872",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2504227_c4sz78k1_2504227/22872",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2504227_c4sz78k1_2504227",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2504227_step22872_24tok_GOOG_2512979/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2504227_c4sz78k1_2504227/22872/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2504227_c4sz78k1_2504227/22872",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/LOBS5/checkpoints/j2504227_c4sz78k1_2504227/22872",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "22482",
                "22872",
                "23258",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/LOBS5/checkpoints/j2504227_c4sz78k1_2504227/22872",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504227_step22872_24tok_GOOG_2512979.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5/checkpoints/j2504227_c4sz78k1_2504227\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2504227_step22872_24tok_GOOG_2512979\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 22872         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2504227_step22872_24tok_GOOG_2512979.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504227_step22872_24tok/scores/scores_uncond_GOOG_j2504227_step22872_24tok_integrated_2512979.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504227_step22872_24tok/scores/scores_uncond_GOOG_j2504227_step22872_24tok_integrated_2512979.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1751882541368693,
        "feature_points": {
          "spread": 0.40264258181495205,
          "ask_cancellation_depth": 0.300429471207872,
          "bid_cancellation_depth": 0.31974270252305714,
          "limit_ask_order_levels": 0.08717071274734096,
          "limit_bid_order_levels": 0.095168315406865,
          "ask_cancellation_levels": 0.15190453573909304,
          "bid_cancellation_levels": 0.1532492278665213,
          "vol_per_min": 0.054140676132083274,
          "ofi": 0.07286488431365391,
          "ofi_up": 0.10666937901533317,
          "ofi_stay": 0.07309640796339492,
          "orderbook_imbalance": 0.10010704365533372,
          "ofi_down": 0.08911534813478088,
          "log_inter_arrival_time": 0.017005590194112113,
          "log_time_to_cancel": 0.4436482187183445,
          "ask_volume_touch": 0.13481835707894796,
          "bid_volume_touch": 0.13810515631167286,
          "ask_volume": 0.11030032134112525,
          "bid_volume": 0.18834849343538987,
          "limit_ask_order_depth": 0.309763425377824,
          "limit_bid_order_depth": 0.3306624878965576
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2504227_step22872_24tok/scores/scores_uncond_GOOG_j2504227_step22872_24tok_integrated_2512979.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.175188,
            "report_record_id": "canonical-scores:37712c45-68fd-81f1-82d7-ffd238f85e0f",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f6-82e1-f1af074b346c",
      "source_row_id": "37712c45-68fd-81f6-82e1-f1af074b346c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f682e1f1af074b346c",
      "run": "gdn-32head-s10800-c500g500-tokens",
      "stock": null,
      "model_size": null,
      "run_family": "gdn",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.558048,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.558048",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.737166,
      "reported_ks_context": 0.680465,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:26Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81f682e1f1af074b346c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:26Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:26Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:26Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.558048,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f6-82e1-f1af074b346c",
        "class": "G20",
        "model_evaluation": "gdn-32head-s10800-c500g500-tokens",
        "wws21_mean": 0.558048,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:26Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c500g500-tokens_GOOG_3437865/data_real",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c500g500-tokens_GOOG_3437865/data_real;/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c500g500-tokens_GOOG_3437855/data_real",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:26Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f682e1f1af074b346c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 977,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273/10800",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c500g500-tokens_GOOG_3437865.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273 (step 10800)",
            "checkpoint_step": 10800,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c500g500-tokens_GOOG_3437865.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/gdn-32head-s10800-c500g500-tokens_GOOG_3437865\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 10800         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.151232+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c500g500-tokens_GOOG_3437865.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_Q1_GDN_scaling/checkpoints/j3422273_me4284ou_3422273\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/gdn-32head-s10800-c500g500-tokens_GOOG_3437865\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 10800         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c500g500-tokens_GOOG_3437865.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_gdn-32head-s10800-c500g500-tokens_GOOG_3437865.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/gdn-32head-s10800-c500g500-tokens_GOOG_3437865"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c500g500-tokens_GOOG_3437865/data_real",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c500g500-tokens_GOOG_3437865/data_real",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/gdn-32head-s10800-c500g500-tokens_GOOG_3437865/data_real",
            "selector": null,
            "source_value": 0.558048,
            "report_record_id": "canonical-scores:37712c45-68fd-81f6-82e1-f1af074b346c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f6-b769-f2407c99f170",
      "source_row_id": "37712c45-68fd-81f6-b769-f2407c99f170",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f6b769f2407c99f170",
      "run": "G10-l10-arm",
      "stock": null,
      "model_size": null,
      "run_family": "G10",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.514499,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.514499",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.391434,
      "reported_ks_context": 0.365807,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:32Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81f6b769f2407c99f170",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:32Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:32Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:32Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.514499,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f6-b769-f2407c99f170",
        "class": "G20",
        "model_evaluation": "G10-l10-arm",
        "wws21_mean": 0.514499,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:32Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-arm_GOOG_3508925/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-arm_GOOG_3508925/data_real.tar;/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-arm_GOOG_3512716/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:32Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 2048 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f6b769f2407c99f170",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 978,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168/133168",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-arm_GOOG_3508925.out",
            "evidence_line": 8,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168 (step 133168)",
            "checkpoint_step": 133168,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-arm_GOOG_3508925.out",
                "evidence_line": 22,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l10-arm_GOOG_3508925\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 2048         --batch_size 64         --checkpoint_step 133168         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.162095+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-arm_GOOG_3508925.out",
            "line": 22,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_G10-l10-vs-l3000/checkpoints/j2823168_zz7t25ww_2823168\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l10-arm_GOOG_3508925\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 2048         --batch_size 64         --checkpoint_step 133168         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-arm_GOOG_3508925.out",
            "line": 14,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_G10-l10-arm_GOOG_3508925.out",
            "line": 19,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/G10-l10-arm_GOOG_3508925"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-arm_GOOG_3508925/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-arm_GOOG_3508925/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/G10-l10-arm_GOOG_3508925/data_real.tar",
            "selector": null,
            "source_value": 0.514499,
            "report_record_id": "canonical-scores:37712c45-68fd-81f6-b769-f2407c99f170",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f7-914f-d1c71090f0e2",
      "source_row_id": "37712c45-68fd-81f7-914f-d1c71090f0e2",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f7914fd1c71090f0e2",
      "run": "N1b-split-s62183-L100sim-fix",
      "stock": null,
      "model_size": null,
      "run_family": "N1b",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.220897,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.220897",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.201488,
      "reported_ks_context": 0.131912,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:42:43Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81f7914fd1c71090f0e2",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:42:43Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:42:43Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:42:43Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.220897,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f7-914f-d1c71090f0e2",
        "class": "G20",
        "model_evaluation": "N1b-split-s62183-L100sim-fix",
        "wws21_mean": 0.220897,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:42:43Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim-fix_GOOG_3234778/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim-fix_GOOG_3234778/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:42:43Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f7914fd1c71090f0e2",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 979,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim-fix_GOOG_3234778.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129 (step 62183)",
            "checkpoint_step": 62183,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim-fix_GOOG_3234778.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1b-split-s62183-L100sim-fix_GOOG_3234778\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 62183         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263186+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.181399+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim-fix_GOOG_3234778.out",
            "line": 21,
            "text": "  Command: /projects/s5e/quant/miniforge3/bin/python3 -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_N1b_vocab_split/checkpoints/j3162129_81xttjij_3162129\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1b-split-s62183-L100sim-fix_GOOG_3234778\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 62183         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim-fix_GOOG_3234778.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_N1b-split-s62183-L100sim-fix_GOOG_3234778.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/N1b-split-s62183-L100sim-fix_GOOG_3234778"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim-fix_GOOG_3234778/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim-fix_GOOG_3234778/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/N1b-split-s62183-L100sim-fix_GOOG_3234778/data_real.tar",
            "selector": null,
            "source_value": 0.220897,
            "report_record_id": "canonical-scores:37712c45-68fd-81f7-914f-d1c71090f0e2",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f7-9ed4-db66aed96aca",
      "source_row_id": "37712c45-68fd-81f7-9ed4-db66aed96aca",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f79ed4db66aed96aca",
      "run": "1tok-75m-lobbench-v2",
      "stock": null,
      "model_size": "75M",
      "run_family": "1tok",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.300679,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.300679",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.258805,
      "reported_ks_context": 0.203253,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:35:16Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81f79ed4db66aed96aca",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:35:16Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:35:16Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:35:16Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.300679,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f7-9ed4-db66aed96aca",
        "class": "G20",
        "model_evaluation": "1tok-75m-lobbench-v2",
        "wws21_mean": 0.300679,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "estimated",
        "date_confidence": "estimated",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:35:16Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_1tok-75m-lobbench-v2/scores/scores_uncond_GOOG_1tok-75m-lobbench-v2_integrated_2859449.pkl",
        "input_path": "",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期及 class 为估计，置信度 high；Recorded full-January sampler, 1024 windows, c=500/g=500 and no fixed index override. All 208 matching command observations linked to established date sets cover 20 days; raw outputs for this run are missing.；评估完成日期未确认；来源原有日期 2026-06-06T21:35:16Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f79ed4db66aed96aca",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 980,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "189804",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_1tok-75m-lobbench-v2_GOOG_2859449.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371 (step 189804)",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371/189804",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "111420",
                "115585",
                "123900",
                "152760",
                "164890",
                "169041",
                "173188",
                "177350",
                "181488",
                "185639",
                "189804",
                "193967",
                "198110",
                "202269",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371/189804",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Source input_path was blank; checkpoint is from the integrated-job log matching the exact score filename job ID. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_1tok-75m-lobbench-v2_GOOG_2859449.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_K2_1tok/checkpoints/j2843371_54c1tkib_2843371\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/1tok-75m-lobbench-v2_GOOG_2859449\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 32         --checkpoint_step 189804         --test_split 1.0 --token_mode 1tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_1tok-75m-lobbench-v2_GOOG_2859449.out",
            "line": 13,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_1tok-75m-lobbench-v2/scores/scores_uncond_GOOG_1tok-75m-lobbench-v2_integrated_2859449.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_1tok-75m-lobbench-v2/scores/scores_uncond_GOOG_1tok-75m-lobbench-v2_integrated_2859449.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.30067918972626767,
        "feature_points": {
          "ofi_up": 0.43950695505601584,
          "bid_cancellation_depth": 0.5719459204952687,
          "ask_volume": 0.11858460030069651,
          "ask_cancellation_depth": 0.5276043936399226,
          "limit_ask_order_levels": 0.24948531043658934,
          "limit_bid_order_levels": 0.2452658885065856,
          "limit_bid_order_depth": 0.02446537551261816,
          "bid_volume": 0.12488936570063108,
          "ask_cancellation_levels": 0.13444211626305094,
          "bid_cancellation_levels": 0.14870033550366002,
          "orderbook_imbalance": 0.18560562715614387,
          "spread": 0.0021970771481247564,
          "vol_per_min": 0.6373671130905103,
          "limit_ask_order_depth": 0.5139825679785154,
          "ofi_down": 0.1502044837517304,
          "log_inter_arrival_time": 0.5634309829531224,
          "bid_volume_touch": 0.190813559144562,
          "log_time_to_cancel": 0.7531584332049741,
          "ofi_stay": 0.28527508653705047,
          "ask_volume_touch": 0.1691153533928239,
          "ofi": 0.27822243847902484
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_1tok-75m-lobbench-v2/scores/scores_uncond_GOOG_1tok-75m-lobbench-v2_integrated_2859449.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.300679,
            "report_record_id": "canonical-scores:37712c45-68fd-81f7-9ed4-db66aed96aca",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f9-9ae7-c137b8bca166",
      "source_row_id": "37712c45-68fd-81f9-9ae7-c137b8bca166",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f99ae7c137b8bca166",
      "run": "pareto-55m-s6446",
      "stock": null,
      "model_size": "55M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.271293,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.271293",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.21148,
      "reported_ks_context": 0.141761,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:33:03Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81f99ae7c137b8bca166",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:33:03Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:33:03Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:33:03Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.271293,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f9-9ae7-c137b8bca166",
        "class": "G20",
        "model_evaluation": "pareto-55m-s6446",
        "wws21_mean": 0.271293,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:33:03Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s6446_GOOG_3253387/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s6446_GOOG_3253387/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:33:03Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f99ae7c137b8bca166",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 981,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916/6446",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s6446_GOOG_3253387.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916 (step 6446)",
            "checkpoint_step": 6446,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s6446_GOOG_3253387.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s6446_GOOG_3253387\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 6446         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.183844+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s6446_GOOG_3253387.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2492916_uu5j082s_2492916\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s6446_GOOG_3253387\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 6446         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s6446_GOOG_3253387.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-55m-s6446_GOOG_3253387.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-55m-s6446_GOOG_3253387"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s6446_GOOG_3253387/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s6446_GOOG_3253387/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-55m-s6446_GOOG_3253387/data_real.tar",
            "selector": null,
            "source_value": 0.271293,
            "report_record_id": "canonical-scores:37712c45-68fd-81f9-9ae7-c137b8bca166",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81f9-9d5c-de388091e50c",
      "source_row_id": "37712c45-68fd-81f9-9d5c-de388091e50c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81f99d5cde388091e50c",
      "run": "H6-MoE-38k",
      "stock": null,
      "model_size": null,
      "run_family": "H6",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.189706,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.189706",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.152644,
      "reported_ks_context": 0.096803,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:28Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81f99d5cde388091e50c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:28Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:28Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:28Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.189706,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81f9-9d5c-de388091e50c",
        "class": "G20",
        "model_evaluation": "H6-MoE-38k",
        "wws21_mean": 0.189706,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:28Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6-MoE-38k/scores/scores_uncond_GOOG_H6-MoE-38k_integrated_2572578.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H6-MoE-38k_GOOG_2572578/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:28Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H6-MoE-38k_GOOG_2572578/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81f99d5cde388091e50c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 982,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "38349",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/H6-MoE-38k_GOOG_2572578/log_rank0.txt",
            "evidence_line": 855,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "13595",
                "16900",
                "25150",
                "28450",
                "29275",
                "30922",
                "31749",
                "32573",
                "33397",
                "34224",
                "35048",
                "35875",
                "36701",
                "37524",
                "38349",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830/38349",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H6-MoE-38k_GOOG_2572578.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H6-moe-scale/checkpoints/j2505830_s5p7vl3x_2505830\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/H6-MoE-38k_GOOG_2572578\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 38349         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_H6-MoE-38k_GOOG_2572578.out",
            "line": 12,
            "text": "Skip infer:  0"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6-MoE-38k/scores/scores_uncond_GOOG_H6-MoE-38k_integrated_2572578.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6-MoE-38k/scores/scores_uncond_GOOG_H6-MoE-38k_integrated_2572578.pkl",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.18970582538728453,
        "feature_points": {
          "spread": 0.22390147449732248,
          "ask_cancellation_depth": 0.38917364136190175,
          "bid_cancellation_depth": 0.4069566458200261,
          "limit_ask_order_levels": 0.06405212290458848,
          "limit_bid_order_levels": 0.07918488797288795,
          "ask_cancellation_levels": 0.1592571355247994,
          "bid_cancellation_levels": 0.12655416684626614,
          "vol_per_min": 0.12163884781951026,
          "ofi": 0.12813274172186054,
          "ofi_up": 0.04174826849517389,
          "ofi_stay": 0.13075979403274204,
          "orderbook_imbalance": 0.09063378424308192,
          "ofi_down": 0.1119072144868238,
          "log_inter_arrival_time": 0.14225636236252845,
          "log_time_to_cancel": 0.5797370744444589,
          "ask_volume_touch": 0.10096601439217893,
          "bid_volume_touch": 0.09434587664619315,
          "ask_volume": 0.07346667979911961,
          "bid_volume": 0.09253477839357471,
          "limit_ask_order_depth": 0.4108501446976757,
          "limit_bid_order_depth": 0.415764676670261
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_H6-MoE-38k/scores/scores_uncond_GOOG_H6-MoE-38k_integrated_2572578.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.189706,
            "report_record_id": "canonical-scores:37712c45-68fd-81f9-9d5c-de388091e50c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81fb-83d7-c019e2cbc452",
      "source_row_id": "37712c45-68fd-81fb-83d7-c019e2cbc452",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81fb83d7c019e2cbc452",
      "run": "phaseB-m3-14m-s4000-d128-b-avg-c500g500",
      "stock": null,
      "model_size": "14M",
      "run_family": "phaseB",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.208309,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.208309",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.208371,
      "reported_ks_context": 0.12097,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:41:55Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81fb83d7c019e2cbc452",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:41:55Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:41:55Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:41:55Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.208309,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81fb-83d7-c019e2cbc452",
        "class": "G20",
        "model_evaluation": "phaseB-m3-14m-s4000-d128-b-avg-c500g500",
        "wws21_mean": 0.208309,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:41:55Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:41:55Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81fb83d7c019e2cbc452",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 983,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800/999999",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942/log_rank0.txt",
            "evidence_line": 628,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800/999999/state.), falling back to direct restore for inference",
            "checkpoint_step": 999999,
            "logged_path": "checkpoints/j3703800_w9cmkmrf_3703800",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703800_w9cmkmrf_3703800\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942.out",
                "evidence_line": 7,
                "evidence_text": "Checkpoint:  checkpoints/j3703800_w9cmkmrf_3703800 (step 999999)"
              },
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942/log_rank0.txt",
                "evidence_line": 630,
                "evidence_text": "[load_checkpoint] Loaded 137 param arrays"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800",
            "availability_checked_at": "2026-09-12T06:13:15.371164+00:00",
            "runtime_logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800/999999/state",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Relative integrated reference resolved by the absolute state path in the same evaluation rank0 restore log. The existing audit prefix migration rule generated the current public candidate, whose selected step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3703800_w9cmkmrf_3703800/999999/state"
          }
        ],
        "note": "The integrated log records a relative checkpoint path. The same evaluation rank0 log records its absolute state path, StandardRestore fallback, then successful direct array loading. The selected step directory exists at the current public path. Weight contents and current loadability were not checked."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"checkpoints/j3703800_w9cmkmrf_3703800\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 999999         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/phaseB-m3-14m-s4000-d128-b-avg-c500g500_GOOG_3713942/data_real.tar",
            "selector": null,
            "source_value": 0.208309,
            "report_record_id": "canonical-scores:37712c45-68fd-81fb-83d7-c019e2cbc452",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81fc-941f-c6c13f7fcf0c",
      "source_row_id": "37712c45-68fd-81fc-941f-c6c13f7fcf0c",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81fc941fc6c13f7fcf0c",
      "run": "pareto-360m-adamw-s5590",
      "stock": null,
      "model_size": "360M",
      "run_family": "pareto",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.305141,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.305141",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.244304,
      "reported_ks_context": 0.165411,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:37:47Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81fc941fc6c13f7fcf0c",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:37:47Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:37:47Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:37:47Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.305141,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81fc-941f-c6c13f7fcf0c",
        "class": "G20",
        "model_evaluation": "pareto-360m-adamw-s5590",
        "wws21_mean": 0.305141,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:37:47Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s5590_GOOG_3253400/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s5590_GOOG_3253400/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:37:47Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 3136 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81fc941fc6c13f7fcf0c",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 984,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s5590_GOOG_3253400.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747 (step 5590)",
            "checkpoint_step": 5590,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s5590_GOOG_3253400.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s5590_GOOG_3253400\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 5590         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263082+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.158732+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s5590_GOOG_3253400.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2689747_ridsve7i_2689747\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s5590_GOOG_3253400\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 3136         --batch_size 64         --checkpoint_step 5590         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s5590_GOOG_3253400.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_pareto-360m-adamw-s5590_GOOG_3253400.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/pareto-360m-adamw-s5590_GOOG_3253400"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s5590_GOOG_3253400/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s5590_GOOG_3253400/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/pareto-360m-adamw-s5590_GOOG_3253400/data_real.tar",
            "selector": null,
            "source_value": 0.305141,
            "report_record_id": "canonical-scores:37712c45-68fd-81fc-941f-c6c13f7fcf0c",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81fd-89d5-d886c70ca3e1",
      "source_row_id": "37712c45-68fd-81fd-89d5-d886c70ca3e1",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81fd89d5d886c70ca3e1",
      "run": "j2496072_ti7ek5z5",
      "stock": null,
      "model_size": null,
      "run_family": "j2496072",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.892951,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.892951",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.712576,
      "reported_ks_context": 0.588158,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-02-27T10:31:40Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2256",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-02-27T10:31:40Z",
      "collection_source": "notion",
      "archive_source_date": "2026-02-27T10:31:40Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-02-27T10:31:40Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.892951,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81fd-89d5-d886c70ca3e1",
        "class": "G20",
        "model_evaluation": "j2496072_ti7ek5z5",
        "wws21_mean": 0.892951,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-02-27T10:31:40Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_shard2_integrated_2643169.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2832163/data_real.tar;/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2643169/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-02-27T10:31:40Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2832163/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81fd89d5d886c70ca3e1",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 985,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "28495",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2643169/log_rank0.txt",
            "evidence_line": 465,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "11337",
                "13247",
                "15160",
                "17061",
                "18965",
                "20871",
                "22773",
                "24680",
                "26591",
                "28495",
                "9435",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072/28495",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Multiple source input candidates were resolved by the exact score filename explicitly listed in the matching integrated-job log. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496072_ti7ek5z5_GOOG_2643169.out",
            "line": 518,
            "text": "  - /lus/lfs1aip2/projects/s5e/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_shard2_integrated_2643169.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 256,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496072_ti7ek5z5_GOOG_2643169.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H4-self-attention/checkpoints/j2496072_ti7ek5z5_2496072\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2496072_ti7ek5z5_GOOG_2643169\"         --n_cond_msgs 256         --n_gen_msgs 500         --n_sequences 1024         --batch_size 16         --checkpoint_step 28495         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496072_ti7ek5z5_GOOG_2643169.out",
            "line": 12,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2496072_ti7ek5z5_GOOG_2643169.out",
            "line": 518,
            "text": "  - /lus/lfs1aip2/projects/s5e/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_shard2_integrated_2643169.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages."
      },
      "score_audit_info": {
        "status": "incomplete_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_shard2_integrated_2643169.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_integrated_2643169.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 20,
        "nonfinite_features": [
          "orderbook_imbalance"
        ],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "reviewed_nonfinite_feature_counts",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2496072_ti7ek5z5/scores/scores_uncond_GOOG_j2496072_ti7ek5z5_integrated_2643169.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.892951,
            "report_record_id": "canonical-scores:37712c45-68fd-81fd-89d5-d886c70ca3e1",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "The reviewed 21-feature score file contains nonfinite feature values. The reported source score is retained; it is not a verified complete finite 21-feature mean. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81fd-a64d-ed2c6e2a2279",
      "source_row_id": "37712c45-68fd-81fd-a64d-ed2c6e2a2279",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81fda64ded2c6e2a2279",
      "run": "adamw-32n-s22705",
      "stock": null,
      "model_size": null,
      "run_family": "adamw",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.174108,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.174108",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.136601,
      "reported_ks_context": 0.089058,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-06-06T21:39:06Z",
      "date_basis": "notion_record_created",
      "date_precision": "timestamp",
      "date_source_field": "createdTime",
      "date_source_reference": "https://app.notion.com/37712c4568fd81fda64ded2c6e2a2279",
      "date_join_status": "none",
      "date_is_evaluation_completion": false,
      "date": "2026-06-06T21:39:06Z",
      "collection_source": "notion",
      "archive_source_date": "2026-06-06T21:39:06Z",
      "archive_date_basis": "notion_record_created",
      "source_date": "2026-06-06T21:39:06Z",
      "source_date_basis": "notion_record_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.174108,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81fd-a64d-ed2c6e2a2279",
        "class": "G20",
        "model_evaluation": "adamw-32n-s22705",
        "wws21_mean": 0.174108,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-06-06T21:39:06Z",
        "source_date_basis": "notion_record_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s22705_GOOG_2869639/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s22705_GOOG_2869639/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-06-06T21:39:06Z（notion_record_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：input_data；该输入档案含 1024 个 message 窗口文件",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81fda64ded2c6e2a2279",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 986,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s22705_GOOG_2869639.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367 (step 22705)",
            "checkpoint_step": 22705,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s22705_GOOG_2869639.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s22705_GOOG_2869639\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 22705         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263003+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.136031+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s22705_GOOG_2869639.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /projects/s5e/lob_pipeline/LOBS5         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_J2_muon_optimizer/checkpoints/j2731367_u5xps1po_2731367\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/adamw-32n-s22705_GOOG_2869639\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 22705         --test_split 1.0 --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s22705_GOOG_2869639.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_adamw-32n-s22705_GOOG_2869639.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/adamw-32n-s22705_GOOG_2869639"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s22705_GOOG_2869639/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s22705_GOOG_2869639/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/adamw-32n-s22705_GOOG_2869639/data_real.tar",
            "selector": null,
            "source_value": 0.174108,
            "report_record_id": "canonical-scores:37712c45-68fd-81fd-a64d-ed2c6e2a2279",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "canonical-scores:37712c45-68fd-81fe-9984-dc11e2b2b241",
      "source_row_id": "37712c45-68fd-81fe-9984-dc11e2b2b241",
      "source_export": "canonical-scores",
      "source_database": "https://app.notion.com/p/37512c4568fd813aa2ded86b8c8ffd8e?pvs=204",
      "source_row_url": "https://app.notion.com/37712c4568fd81fe9984dc11e2b2b241",
      "run": "j2514440_step135458_24tok_perftest",
      "stock": null,
      "model_size": null,
      "run_family": "j2514440",
      "score_families": [
        "uncond"
      ],
      "mean_ws21": 0.134382,
      "source_value_field": "Wasserstein",
      "source_value_text": "0.134382",
      "numeric_field_exact_decimal_roundtrip": true,
      "reported_l1_context": 0.117824,
      "reported_ks_context": 0.075862,
      "n_batches": null,
      "inventory": null,
      "recorded_at": "2026-03-01T08:19:43Z",
      "date_basis": "sheet_recorded_run_created",
      "date_precision": "timestamp",
      "date_source_field": "timestamp",
      "date_source_reference": "work/sheets/lobs5-experiments-verified.xlsx#wandb_runs!row=2150",
      "date_join_status": "matched",
      "date_is_evaluation_completion": false,
      "date": "2026-03-01T08:19:43Z",
      "collection_source": "notion",
      "archive_source_date": "2026-03-01T08:19:43Z",
      "archive_date_basis": "sheet_recorded_run_created",
      "source_date": "2026-03-01T08:19:43Z",
      "source_date_basis": "sheet_recorded_run_created",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.134382,
      "g20_classification": {
        "id": "canonical-scores:37712c45-68fd-81fe-9984-dc11e2b2b241",
        "class": "G20",
        "model_evaluation": "j2514440_step135458_24tok_perftest",
        "wws21_mean": 0.134382,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-01T08:19:43Z",
        "source_date_basis": "sheet_recorded_run_created",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440_step135458_24tok_perftest/scores/scores_uncond_GOOG_j2514440_step135458_24tok_perftest_shard8_integrated_2540822.pkl",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/data_real.tar",
        "path_type": "score_file",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-01T08:19:43Z（sheet_recorded_run_created）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；path 类型：score_file；该输入档案含 1024 个 message 窗口文件；input_path=/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/data_real.tar",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://app.notion.com/37712c4568fd81fe9984dc11e2b2b241",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 987,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "checkpoint_step": "135458",
            "path_kind": "directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "historical_root": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/log_rank0.txt",
            "evidence_line": 633,
            "evidence_text": "[load_checkpoint] StandardRestore failed (Checkpoint structure file does not exist at /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458/state.), falling back to direct restore for inference",
            "migration_evidence": {
              "path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
              "lines": [
                {
                  "line": 15,
                  "text": "def normalize(path):"
                },
                {
                  "line": 16,
                  "text": "    path = str(path)"
                },
                {
                  "line": 17,
                  "text": "    if path.startswith('../../'):"
                },
                {
                  "line": 18,
                  "text": "        path = '/projects/public/u6gb/' + path[6:]"
                },
                {
                  "line": 19,
                  "text": "    for old, new in [('/lus/lfs1aip2/projects/public/', '/projects/public/'),"
                },
                {
                  "line": 20,
                  "text": "                     ('/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 21,
                  "text": "                     ('/lus/lfs1aip2/projects/u6gb/public/', '/projects/public/u6gb/'),"
                },
                {
                  "line": 22,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 23,
                  "text": "                     ('/projects/s5e/lob_pipeline/', str(BASES[0]) + '/'),"
                },
                {
                  "line": 24,
                  "text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
                },
                {
                  "line": 25,
                  "text": "                     ('/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/')]:"
                },
                {
                  "line": 26,
                  "text": "        if path.startswith(old):"
                },
                {
                  "line": 27,
                  "text": "            path = new + path[len(old):]"
                },
                {
                  "line": 28,
                  "text": "    return path"
                }
              ]
            },
            "historical_stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "not_found"
            },
            "stat_evidence": {
              "host": "u6gb.aip2.isambard",
              "checked_at": "2026-09-12T06:09:46.611682+00:00",
              "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
              "availability": "exists",
              "entries": [
                "_CHECKPOINT_METADATA",
                "metadata",
                "state"
              ],
              "root_entries": [
                "131224",
                "133341",
                "135458",
                "metadata"
              ]
            },
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440/135458",
            "availability": "exists",
            "path_scope": "checkpoint_step",
            "resolution": "recorded_prefix_mapping_and_exact_run_step_directory"
          }
        ],
        "note": "Scoring skipped inference and explicitly reused prior outputs. Checkpoint comes from the reused inference runtime log; the rescoring checkpoint header was not treated as runtime evidence. Current path uses the recorded historical-to-public prefix mapping; the exact run/step directory and checkpoint entries were verified. Weight content and loadability were not tested.",
        "score_link_evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/bench_data/j2514440_step135458_24tok/GOOG/2023/data_real",
            "operation": "readlink",
            "text": "/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/data_real"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/bench_data/j2514440_step135458_24tok/GOOG/2023/data_gen",
            "operation": "readlink",
            "text": "/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/data_gen"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_perftest_GOOG_2540822.out",
            "line": 12,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_perftest_GOOG_2540822.out",
            "line": 16,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/bench_data/j2514440_step135458_24tok/GOOG/2023"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_perftest_GOOG_2540822.out",
            "line": 506,
            "text": "  - /lus/lfs1aip2/projects/s5e/lob_pipeline/results_j2514440_step135458_24tok_perftest/scores/scores_uncond_GOOG_j2514440_step135458_24tok_perftest_shard8_integrated_2540822.pkl"
          }
        ]
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_GOOG_2518009.out",
            "line": 18,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_H1-scaling-law/checkpoints/j2514440_bkotgtm5_2514440\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 135458         --test_split 1.0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_GOOG_2518009.out",
            "line": 12,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/bench_data/j2514440_step135458_24tok/GOOG/2023/data_real",
            "text": "/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/data_real"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/bench_data/j2514440_step135458_24tok/GOOG/2023/data_gen",
            "text": "/lus/lfs1aip2/projects/s5e/lob_pipeline/LOBS5/inference_results/j2514440_step135458_24tok_GOOG_2518009/data_gen"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_perftest_GOOG_2540822.out",
            "line": 12,
            "text": "Skip infer:  1"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_perftest_GOOG_2540822.out",
            "line": 16,
            "text": "  Using existing: /lus/lfs1aip2/projects/s5e/lob_pipeline/bench_data/j2514440_step135458_24tok/GOOG/2023"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_j2514440_step135458_24tok_perftest_GOOG_2540822.out",
            "line": 506,
            "text": "  - /lus/lfs1aip2/projects/s5e/lob_pipeline/results_j2514440_step135458_24tok_perftest/scores/scores_uncond_GOOG_j2514440_step135458_24tok_perftest_shard8_integrated_2540822.pkl"
          }
        ],
        "note": "Evaluation conditioning and requested generation counts are the explicit --n_cond_msgs and --n_gen_msgs arguments, in messages. This score reused prior inference outputs; the displayed counts come from the original inference command. The recorded source-output reuse chain is retained in the evidence. Benchmark staging data_real/data_gen symlink targets were checked to connect the reused benchmark directory to the original inference run."
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440_step135458_24tok_perftest/scores/scores_uncond_GOOG_j2514440_step135458_24tok_perftest_shard8_integrated_2540822.pkl",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440_step135458_24tok_perftest/scores/scores_uncond_GOOG_j2514440_step135458_24tok_perftest_integrated_2540822.pkl",
        "path_resolution": "reviewed_merged_counterpart",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.134381929644474,
        "feature_points": {
          "spread": 0.17825713495974982,
          "ask_cancellation_depth": 0.3187137685929295,
          "bid_cancellation_depth": 0.325966649826433,
          "limit_ask_order_levels": 0.06287941660018229,
          "limit_bid_order_levels": 0.07466590898520135,
          "ask_cancellation_levels": 0.08212132026297445,
          "bid_cancellation_levels": 0.07156062942363556,
          "vol_per_min": 0.07259737054661505,
          "ofi": 0.04909453568875983,
          "ofi_up": 0.029500603148470632,
          "ofi_stay": 0.0503617331023333,
          "orderbook_imbalance": 0.11943404264139905,
          "ofi_down": 0.042019703752596636,
          "log_inter_arrival_time": 0.06570551752048691,
          "log_time_to_cancel": 0.3674359457736012,
          "ask_volume_touch": 0.06621965099994667,
          "bid_volume_touch": 0.05541910785409989,
          "ask_volume": 0.07013180473983316,
          "bid_volume": 0.047048056355463816,
          "limit_ask_order_depth": 0.3285509710442277,
          "limit_bid_order_depth": 0.3443366507150146
        },
        "evidence": [
          {
            "kind": "saved_raw_feature_points",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/results_j2514440_step135458_24tok_perftest/scores/scores_uncond_GOOG_j2514440_step135458_24tok_perftest_integrated_2540822.pkl",
            "selector": "tuple[0]/*/wasserstein[0]",
            "source_value": 0.134382,
            "report_record_id": "canonical-scores:37712c45-68fd-81fe-9984-dc11e2b2b241",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "hf-board-a-m3-34m-eggroll-4n-p256-s300-v3",
      "mean_ws21": 0.1298,
      "date": "2026-03-29",
      "date_basis": "hf_recorded_training_start",
      "timestamp": "2026-03-29T18:47:22Z",
      "source": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/4d2edf9f72fe03a540581b4f9a3b17e9537d8fda/data/r1_mamba3_lobbench.json#board_a[22].ws",
      "model": "m3-34m-eggroll-4n-P256-s300-v3",
      "run_id": "mquj2ymf",
      "job_id": null,
      "step": 300,
      "task": null,
      "protocol": "c250g250",
      "metric_basis": "HF board metadata explicitly states arithmetic mean over 21 Wasserstein feature scores.",
      "source_record": {
        "model": "m3-34m-eggroll-4n-P256-s300-v3",
        "variant": "ablation",
        "size": "34m",
        "steps": 300,
        "protocol": "c250g250",
        "params": 33610439,
        "ws": 0.1298,
        "ks": 0.1054,
        "l1": 0.1981,
        "n_features": 21,
        "wandb_id": "mquj2ymf",
        "start_date": "2026-03-29",
        "start_ts": "2026-03-29T18:47:22Z",
        "end_date": "2026-03-29",
        "end_ts": "2026-03-29T18:49:13Z",
        "runtime_h": 3.28,
        "run_state": "crashed",
        "n_stocks": 8,
        "n_years": 4,
        "data_range": "2022–2025",
        "wandb_url": "https://wandb.ai/oxford-lob/mamba3-siso-scaling-law/runs/mquj2ymf",
        "ckpt_time_estimated": true,
        "rank": 23
      },
      "date_note": "Some ablations reuse the source checkpoint run timestamp; source-recorded training date, not evaluation completion.",
      "collection_source": "sheets_and_previous_hf",
      "archive_source_date": "2026-03-29",
      "archive_date_basis": "hf_recorded_training_start",
      "source_date": "2026-03-29",
      "source_date_basis": "hf_recorded_training_start",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 0.1298,
      "g20_classification": {
        "id": "hf-board-a-m3-34m-eggroll-4n-p256-s300-v3",
        "class": "G20",
        "model_evaluation": "m3-34m-eggroll-4n-P256-s300-v3",
        "wws21_mean": 0.1298,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-29",
        "source_date_basis": "hf_recorded_training_start",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-29（hf_recorded_training_start）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；HF board metadata explicitly states arithmetic mean over 21 Wasserstein feature scores.；path 类型：input_data；protocol=c250g250；该输入档案含 1024 个 message 窗口文件",
        "protocol": "c250g250",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/4d2edf9f72fe03a540581b4f9a3b17e9537d8fda/data/r1_mamba3_lobbench.json#board_a[22].ws",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 988,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints",
            "availability": "not_found",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints (step 300)",
            "checkpoint_step": 300,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 300         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:08:05.263181+00:00",
            "current_path_candidate": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints",
            "current_path_candidate_availability": "not_found",
            "current_path_candidate_checked_at": "2026-09-12T06:10:01.181280+00:00",
            "availability_scope": "logged_checkpoint_directory",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The logged directory and the mapped current directory were both absent at verification; no checkpoint filename was invented."
      },
      "length_info": {
        "context_length": 500,
        "generation_length": 500,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_S1_rollout/eggroll_output/j3642158/eggroll_checkpoints\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607\"         --n_cond_msgs 500         --n_gen_msgs 500         --n_sequences 1024         --batch_size 64         --checkpoint_step 300         --test_split 1.0 --token_mode 26tok --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607.out",
            "line": 18,
            "text": "  Output: /lus/lfs1aip2/projects/s5e/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-34m-eggroll-4n-P256-s300-v3_GOOG_3661607/data_real.tar",
            "selector": null,
            "source_value": 0.1298,
            "report_record_id": "hf-board-a-m3-34m-eggroll-4n-p256-s300-v3",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "hf-board-a-m3-paperbase-pw8u-s46050-c250g250-absmax8k",
      "mean_ws21": 1.1345,
      "date": "2026-03-28",
      "date_basis": "hf_recorded_training_start",
      "timestamp": "2026-03-28T02:04:31Z",
      "source": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/4d2edf9f72fe03a540581b4f9a3b17e9537d8fda/data/r1_mamba3_lobbench.json#board_a[29].ws",
      "model": "m3-paperbase-pw8u-s46050-c250g250-absmax8k",
      "run_id": "pw8u0edj",
      "job_id": null,
      "step": 46050,
      "task": null,
      "protocol": "c250g250",
      "metric_basis": "HF board metadata explicitly states arithmetic mean over 21 Wasserstein feature scores.",
      "source_record": {
        "model": "m3-paperbase-pw8u-s46050-c250g250-absmax8k",
        "variant": "ablation",
        "size": "78m",
        "steps": 46050,
        "protocol": "c250g250",
        "params": 78539423,
        "ws": 1.1345,
        "ks": 0.6562,
        "l1": 0.781,
        "n_features": 21,
        "wandb_id": "pw8u0edj",
        "start_date": "2026-03-28",
        "start_ts": "2026-03-28T02:04:31Z",
        "end_date": "2026-03-28",
        "end_ts": "2026-03-28T10:05:43Z",
        "runtime_h": 8.02,
        "run_state": "finished",
        "n_stocks": 8,
        "n_years": 4,
        "data_range": "2022–2025",
        "wandb_url": "https://wandb.ai/oxford-lob/mamba3/runs/pw8u0edj",
        "ckpt_time_estimated": true,
        "rank": 30
      },
      "date_note": "Some ablations reuse the source checkpoint run timestamp; source-recorded training date, not evaluation completion.",
      "collection_source": "sheets_and_previous_hf",
      "archive_source_date": "2026-03-28",
      "archive_date_basis": "hf_recorded_training_start",
      "source_date": "2026-03-28",
      "source_date_basis": "hf_recorded_training_start",
      "evaluation_completed_at": null,
      "evaluated_at": null,
      "class": "G20",
      "wws21_mean": 1.1345,
      "g20_classification": {
        "id": "hf-board-a-m3-paperbase-pw8u-s46050-c250g250-absmax8k",
        "class": "G20",
        "model_evaluation": "m3-paperbase-pw8u-s46050-c250g250-absmax8k",
        "wws21_mean": 1.1345,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-03-28",
        "source_date_basis": "hf_recorded_training_start",
        "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
        "input_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
        "path_type": "input_data",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-03-28（hf_recorded_training_start）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；HF board metadata explicitly states arithmetic mean over 21 Wasserstein feature scores.；path 类型：input_data；protocol=c250g250；该输入档案含 3136 个 message 窗口文件",
        "protocol": "c250g250",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/4d2edf9f72fe03a540581b4f9a3b17e9537d8fda/data/r1_mamba3_lobbench.json#board_a[29].ws",
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 989,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629/46050",
            "availability": "exists",
            "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
            "evidence_line": 7,
            "evidence_text": "Checkpoint:  /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629 (step 46050)",
            "checkpoint_step": 46050,
            "logged_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629",
            "additional_evidence": [
              {
                "evidence_path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
                "evidence_line": 21,
                "evidence_text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 3136         --batch_size 64         --checkpoint_step 46050         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136_absmax8000_as_s500.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
              },
              {
                "evidence_path": "/projects/public/u6gb/agent_outputs/leaderboard_dataset_classes_20260912/recover_january_days.py",
                "evidence_line": 24,
                "evidence_text": "                     ('/lus/lfs1aip2/projects/s5e/quant/', '/projects/public/s5e/quant_team/quant/'),"
              }
            ],
            "logged_availability": "not_found",
            "availability_checked_at": "2026-09-12T06:10:01.166575+00:00",
            "path_kind": "directory",
            "checkpoint_directory": "/projects/public/s5e/quant_team/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629",
            "availability_scope": "selected_checkpoint_step_directory",
            "path_resolution": "Logged directory mapped with the existing audit prefix migration rule; the explicitly requested numbered step directory was independently stat-confirmed.",
            "historical_path": "/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629"
          }
        ],
        "note": "Explicit checkpoint directory and step matched between the integrated header and inference command. The selected step directory exists at the current public path. Weight contents and loadability were not checked."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
            "line": 21,
            "text": "  Command: /lus/lfs1aip2/projects/s5e/quant/miniforge3/envs/lobs5/bin/python -u run_inference.py         --model_code_dir /lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3         --stock GOOG         --ckpt_path \"/lus/lfs1aip2/projects/s5e/quant/AlphaTrade/experiments/exp_R1_Mamba3/checkpoints/j3417629_pw8u0edj_3417629\"         --data_dir \"/lus/lfs1aip2/projects/s5e/lob_pipeline/data/GOOG_jan2026\"         --save_dir \"/projects/s5e/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744\"         --n_cond_msgs 250         --n_gen_msgs 250         --n_sequences 3136         --batch_size 64         --checkpoint_step 46050         --test_split 1.0 --sample_indices_file \"/lus/lfs1aip2/projects/s5e/lob_pipeline/pipeline/sample_indices/GOOG_3136_absmax8000_as_s500.txt\" --wide_book_dir \"/lus/lfs1aip2/projects/s5e/lob_preproc_l100/GOOG\" --wide_levels 100"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
            "line": 13,
            "text": "Skip infer:  0"
          },
          {
            "path": "/projects/public/s5e/quant_team/lob_pipeline/logs/integrated_mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744.out",
            "line": 18,
            "text": "  Output: /projects/s5e/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744"
          }
        ],
        "note": "Evaluation inference settings explicitly supplied as --n_cond_msgs and --n_gen_msgs in the same run command, with Skip infer: 0. These are conditioning and requested generation message counts, not training token context or a rolling decoder window; raw output arrays were not inspected."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
        "display_path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/s5e/quant_team/lob_pipeline/inference_results/mamba3-paperbase-pw8u-s46050-c250g250-absmax8k_GOOG_4237744/data_real.tar",
            "selector": null,
            "source_value": 1.1345,
            "report_record_id": "hf-board-a-m3-paperbase-pw8u-s46050-c250g250-absmax8k",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "icaif-recurrent-vs-transformer-lob-ws-21-mamba-3-vs-transformer-baseline",
      "experiment_slug": "icaif-recurrent-vs-transformer-lob",
      "label": "WS-21 (Mamba-3 vs Transformer) (reported baseline)",
      "mean_ws21": 0.31,
      "reported_value": 0.31,
      "metric_name": "WS-21 (Mamba-3 vs Transformer)",
      "metric_kind": "unconditional_ws21",
      "aggregation_level": "reported aggregate point estimate",
      "protocol_scope": "three training seeds; parameter-matched Mamba-3 vs Transformer; GOOG January-2026 generation/LOB-Bench",
      "record_role": "baseline",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment",
      "date_precision": "month",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
      "date": "2026-08",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08",
      "archive_date_basis": "registry metadata date_experiment",
      "archive_evaluated_at": "2026-08",
      "source_date": "2026-08",
      "source_date_basis": "registry metadata date_experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.31,
      "g20_classification": {
        "id": "icaif-recurrent-vs-transformer-lob-ws-21-mamba-3-vs-transformer-baseline",
        "class": "G20",
        "model_evaluation": "WS-21 (Mamba-3 vs Transformer) (reported baseline)",
        "wws21_mean": 0.31,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08",
        "source_date_basis": "registry metadata date_experiment",
        "path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08（registry metadata date_experiment）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；reported aggregate point estimate；three training seeds; parameter-matched Mamba-3 vs Transformer; GOOG January-2026 generation/LOB-Bench；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 990,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "This reported score is a three-seed architecture aggregate, not a single checkpoint. The canonical metadata and report specify seeds 5, 42 and 137 and terminal step 30000, but record no checkpoint filesystem paths or upstream run links. No model paths were inferred from scores or names.",
        "completeness": "unresolved",
        "expected_component_count": 3,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seeds": [
          5,
          42,
          137
        ],
        "source_metadata_path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
        "source_report_path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/result.md",
        "source_report_step_evidence_line": 243,
        "source_report_step_evidence_text": "- **Checkpoint**: terminal 30,000 updates for every architecture (checkpoints written every 1,000)."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/result.md",
            "line": 96,
            "text": "**Primary evaluation.** **3,136 windows** drawn uniformly without replacement (selection seed 42) from **226,002 eligible, non-overlapping 500-message windows** across all **20 January 2026 regular-session days**. Windows cannot cross the daily close, so sampling weights each day by its eligible-window count. Each sample = **250 real conditioning messages + 250 generated messages**; generation samples the full syntax-masked categorical at **temperature 1.0**, seed 2026 with deterministic rank offsets. All architectures and seeds share the same windows and sampling settings. The 500-message total matches training length, so the Transformer is never asked to extrapolate its context."
          },
          {
            "path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/result.md",
            "line": 244,
            "text": "- **Eval panel**: 3,136 windows from 226,002 eligible non-overlapping 500-message windows over 20 Jan-2026 sessions; 250 conditioning + 250 generated; temperature 1.0."
          }
        ],
        "note": "The architecture score aggregates three model seeds under explicitly shared evaluation windows and sampling settings: 250 conditioning plus 250 generated messages. The 500-message / 13000-token training sequence is the combined window length, not the conditioning length.",
        "evidence_scope": "Common evaluation protocol explicitly covers all architectures and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
            "selector": null,
            "source_value": 0.31,
            "report_record_id": "icaif-recurrent-vs-transformer-lob-ws-21-mamba-3-vs-transformer-baseline",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "icaif-recurrent-vs-transformer-lob-ws-21-mamba-3-vs-transformer-primary",
      "experiment_slug": "icaif-recurrent-vs-transformer-lob",
      "label": "WS-21 (Mamba-3 vs Transformer) (reported value)",
      "mean_ws21": 0.223,
      "reported_value": 0.223,
      "metric_name": "WS-21 (Mamba-3 vs Transformer)",
      "metric_kind": "unconditional_ws21",
      "aggregation_level": "reported aggregate point estimate",
      "protocol_scope": "three training seeds; parameter-matched Mamba-3 vs Transformer; GOOG January-2026 generation/LOB-Bench",
      "record_role": "primary",
      "evaluated_at": null,
      "date_basis": "registry metadata date_experiment",
      "date_precision": "month",
      "source_artifact": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
      "date": "2026-08",
      "collection_source": "current_experiment_registry",
      "archive_source_date": "2026-08",
      "archive_date_basis": "registry metadata date_experiment",
      "archive_evaluated_at": "2026-08",
      "source_date": "2026-08",
      "source_date_basis": "registry metadata date_experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.223,
      "g20_classification": {
        "id": "icaif-recurrent-vs-transformer-lob-ws-21-mamba-3-vs-transformer-primary",
        "class": "G20",
        "model_evaluation": "WS-21 (Mamba-3 vs Transformer) (reported value)",
        "wws21_mean": 0.223,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "documented",
        "date_confidence": "documented",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-08",
        "source_date_basis": "registry metadata date_experiment",
        "path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
        "input_path": "",
        "path_type": "source_artifact",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；日期依据明确的评测协议或报告；评估完成日期未确认；来源原有日期 2026-08（registry metadata date_experiment）；不作为评估完成时间；保留榜单原报分数；本表未核验全部 21 个点估计；reported aggregate point estimate；three training seeds; parameter-matched Mamba-3 vs Transformer; GOOG January-2026 generation/LOB-Bench；path 类型：source_artifact",
        "protocol": "",
        "score_status": "source_reported_not_revalidated_21_points",
        "source_url": null,
        "source_selector": null,
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 991,
        "full21_provenance": {
          "feature_count": null,
          "validation": "source-reported score retained; all 21 per-feature point estimates unavailable in public record",
          "evidence": [],
          "source_artifact": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "not_found",
        "checkpoints": [],
        "note": "This reported score is a three-seed architecture aggregate, not a single checkpoint. The canonical metadata and report specify seeds 5, 42 and 137 and terminal step 30000, but record no checkpoint filesystem paths or upstream run links. No model paths were inferred from scores or names.",
        "completeness": "unresolved",
        "expected_component_count": 3,
        "resolved_component_count": 0,
        "checkpoint_step": 30000,
        "model_seeds": [
          5,
          42,
          137
        ],
        "source_metadata_path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
        "source_report_path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/result.md",
        "source_report_step_evidence_line": 243,
        "source_report_step_evidence_text": "- **Checkpoint**: terminal 30,000 updates for every architecture (checkpoints written every 1,000)."
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/result.md",
            "line": 96,
            "text": "**Primary evaluation.** **3,136 windows** drawn uniformly without replacement (selection seed 42) from **226,002 eligible, non-overlapping 500-message windows** across all **20 January 2026 regular-session days**. Windows cannot cross the daily close, so sampling weights each day by its eligible-window count. Each sample = **250 real conditioning messages + 250 generated messages**; generation samples the full syntax-masked categorical at **temperature 1.0**, seed 2026 with deterministic rank offsets. All architectures and seeds share the same windows and sampling settings. The 500-message total matches training length, so the Transformer is never asked to extrapolate its context."
          },
          {
            "path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/result.md",
            "line": 244,
            "text": "- **Eval panel**: 3,136 windows from 226,002 eligible non-overlapping 500-message windows over 20 Jan-2026 sessions; 250 conditioning + 250 generated; temperature 1.0."
          }
        ],
        "note": "The architecture score aggregates three model seeds under explicitly shared evaluation windows and sampling settings: 250 conditioning plus 250 generated messages. The 500-message / 13000-token training sequence is the combined window length, not the conditioning length.",
        "evidence_scope": "Common evaluation protocol explicitly covers all architectures and component seeds."
      },
      "score_audit_info": {
        "status": "source_only",
        "original_path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
        "display_path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
        "path_resolution": "original_recorded_path",
        "feature_count": null,
        "finite_feature_count": null,
        "nonfinite_features": [],
        "recomputed_mean": null,
        "feature_points": null,
        "evidence": [
          {
            "kind": "independent_review_source_value",
            "path": "/projects/public/u6gb/results/content/pre-training/icaif-recurrent-vs-transformer-lob/metadata.json",
            "selector": null,
            "source_value": 0.223,
            "report_record_id": "icaif-recurrent-vs-transformer-lob-ws-21-mamba-3-vs-transformer-primary",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8"
          }
        ],
        "note": "Source value checked against saved evidence; a complete finite 21-feature mean was not recomputed. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-field10k-large-s20000",
      "model": "Mamba3 field10k large seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.1650404860489859,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-field10k-large-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.09257208095342083,
        "ask_cancellation_ticks": 0.09219895657221272,
        "ask_volume": 0.036601001682208145,
        "ask_volume_touch": 0.036393543843130254,
        "bid_cancellation_depth": 0.379447912099134,
        "bid_cancellation_ticks": 0.37913092856644615,
        "bid_volume": 0.029646039501205916,
        "bid_volume_touch": 0.03678223329765373,
        "limit_ask_order_depth": 0.10005190598057773,
        "limit_ask_order_ticks": 0.10003968704374822,
        "limit_bid_order_depth": 0.38450067830041007,
        "limit_bid_order_ticks": 0.3837401999029051,
        "log_inter_arrival_time": 0.12502685743554298,
        "log_time_to_cancel": 0.36598700073959933,
        "ofi": 0.19296710581142829,
        "ofi_down": 0.18407938257355078,
        "ofi_stay": 0.19480142709733567,
        "ofi_up": 0.14989255310678817,
        "orderbook_imbalance": 0.08058745862819537,
        "spread": 0.07867682256043629,
        "vol_per_min": 0.04272643133277448
      },
      "evaluation_key": "814915de987b8980ca5be940d17a4fdd01a86060211e1c8f09e12ce25f499e82",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.1650404860489859,
      "g20_classification": {
        "id": "mag7-field10k-large-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 field10k large seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.1650404860489859,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 992,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.16504048604898594,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-field10k-large-s20000.json"
            }
          ],
          "evaluation_key": "814915de987b8980ca5be940d17a4fdd01a86060211e1c8f09e12ce25f499e82",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_large_aligned_seed42/attempts/attempt-0001/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_large_aligned_seed42/attempts/attempt-0001/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_large_aligned_seed42/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "45c58f2d2e374801a1903c428b2460f8cf05c7582578caf15650da5d799e42b9",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_large_aligned_seed42/attempts/attempt-0001/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1650404860489859,
        "feature_points": {
          "ask_cancellation_depth": 0.09257208095342083,
          "ask_cancellation_ticks": 0.09219895657221272,
          "ask_volume": 0.036601001682208145,
          "ask_volume_touch": 0.036393543843130254,
          "bid_cancellation_depth": 0.379447912099134,
          "bid_cancellation_ticks": 0.37913092856644615,
          "bid_volume": 0.029646039501205916,
          "bid_volume_touch": 0.03678223329765373,
          "limit_ask_order_depth": 0.10005190598057773,
          "limit_ask_order_ticks": 0.10003968704374822,
          "limit_bid_order_depth": 0.38450067830041007,
          "limit_bid_order_ticks": 0.3837401999029051,
          "log_inter_arrival_time": 0.12502685743554298,
          "log_time_to_cancel": 0.36598700073959933,
          "ofi": 0.19296710581142829,
          "ofi_down": 0.18407938257355078,
          "ofi_stay": 0.19480142709733567,
          "ofi_up": 0.14989255310678817,
          "orderbook_imbalance": 0.08058745862819537,
          "spread": 0.07867682256043629,
          "vol_per_min": 0.04272643133277448
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-9/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.1650404860489859,
            "report_record_id": "mag7-field10k-large-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-field10k-large-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-field10k-medium-s20000",
      "model": "Mamba3 field10k medium seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.18381917026588485,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-field10k-medium-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.15235446028197522,
        "ask_cancellation_ticks": 0.1520403008585405,
        "ask_volume": 0.029524409408564815,
        "ask_volume_touch": 0.05628458092285514,
        "bid_cancellation_depth": 0.33900558913457923,
        "bid_cancellation_ticks": 0.33894236381388165,
        "bid_volume": 0.015838858578025385,
        "bid_volume_touch": 0.053769137156454136,
        "limit_ask_order_depth": 0.26015009345119366,
        "limit_ask_order_ticks": 0.26016185114409773,
        "limit_bid_order_depth": 0.3739618088591802,
        "limit_bid_order_ticks": 0.37311976113609757,
        "log_inter_arrival_time": 0.13534864958027748,
        "log_time_to_cancel": 0.38347790781924035,
        "ofi": 0.20386674576037583,
        "ofi_down": 0.14754189596319506,
        "ofi_stay": 0.2054971942664178,
        "ofi_up": 0.15732903573592305,
        "orderbook_imbalance": 0.08726210238249361,
        "spread": 0.10600331885810807,
        "vol_per_min": 0.028722510472105157
      },
      "evaluation_key": "e51b7919c9bdcd6b17f4109572a9ed67749a1349cf50d197b10aeee9c0869650",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.18381917026588485,
      "g20_classification": {
        "id": "mag7-field10k-medium-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 field10k medium seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.18381917026588485,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 993,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.18381917026588485,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-field10k-medium-s20000.json"
            }
          ],
          "evaluation_key": "e51b7919c9bdcd6b17f4109572a9ed67749a1349cf50d197b10aeee9c0869650",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_medium_aligned_seed42/attempts/attempt-0001/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_medium_aligned_seed42/attempts/attempt-0001/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_medium_aligned_seed42/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "71d7aa60deac97ae8eec11324c2558ea8e09c1efa7949e9ca5d9a01e8eb8c1b3",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_medium_aligned_seed42/attempts/attempt-0001/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.18381917026588485,
        "feature_points": {
          "ask_cancellation_depth": 0.15235446028197522,
          "ask_cancellation_ticks": 0.1520403008585405,
          "ask_volume": 0.029524409408564815,
          "ask_volume_touch": 0.05628458092285514,
          "bid_cancellation_depth": 0.33900558913457923,
          "bid_cancellation_ticks": 0.33894236381388165,
          "bid_volume": 0.015838858578025385,
          "bid_volume_touch": 0.053769137156454136,
          "limit_ask_order_depth": 0.26015009345119366,
          "limit_ask_order_ticks": 0.26016185114409773,
          "limit_bid_order_depth": 0.3739618088591802,
          "limit_bid_order_ticks": 0.37311976113609757,
          "log_inter_arrival_time": 0.13534864958027748,
          "log_time_to_cancel": 0.38347790781924035,
          "ofi": 0.20386674576037583,
          "ofi_down": 0.14754189596319506,
          "ofi_stay": 0.2054971942664178,
          "ofi_up": 0.15732903573592305,
          "orderbook_imbalance": 0.08726210238249361,
          "spread": 0.10600331885810807,
          "vol_per_min": 0.028722510472105157
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-5/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.18381917026588485,
            "report_record_id": "mag7-field10k-medium-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-field10k-medium-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-field10k-small-s20000",
      "model": "Mamba3 field10k small seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.19232152785569975,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-field10k-small-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.16078483285537223,
        "ask_cancellation_ticks": 0.16034055266294467,
        "ask_volume": 0.09541223691036332,
        "ask_volume_touch": 0.05560360135054905,
        "bid_cancellation_depth": 0.36614365151825323,
        "bid_cancellation_ticks": 0.3660282480702693,
        "bid_volume": 0.07380460650019885,
        "bid_volume_touch": 0.06149689618718125,
        "limit_ask_order_depth": 0.11817116504687425,
        "limit_ask_order_ticks": 0.11819818138644025,
        "limit_bid_order_depth": 0.36128147931077725,
        "limit_bid_order_ticks": 0.3605375514686384,
        "log_inter_arrival_time": 0.10949805827840052,
        "log_time_to_cancel": 0.23188787376605208,
        "ofi": 0.2983454605411607,
        "ofi_down": 0.269833749153752,
        "ofi_stay": 0.29991145969477195,
        "ofi_up": 0.23185605165781892,
        "orderbook_imbalance": 0.0899843723309798,
        "spread": 0.18953516001905793,
        "vol_per_min": 0.020096896259838903
      },
      "evaluation_key": "99328fd9dccd62df406a9f7bccb0f519dee6c3c69dc49d00cb4b5a221ea4cf38",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.19232152785569975,
      "g20_classification": {
        "id": "mag7-field10k-small-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 field10k small seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.19232152785569975,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 994,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.1923215278556998,
          "absolute_difference": 5.551115123125783e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-field10k-small-s20000.json"
            }
          ],
          "evaluation_key": "99328fd9dccd62df406a9f7bccb0f519dee6c3c69dc49d00cb4b5a221ea4cf38",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_small_aligned_seed42/attempts/attempt-0001/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_small_aligned_seed42/attempts/attempt-0001/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_small_aligned_seed42/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "502bf43bc8a69b3c57b05c3b42875b2692fd962e7f5f0ed8908e875f16f795f8",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_field10k_small_aligned_seed42/attempts/attempt-0001/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19232152785569975,
        "feature_points": {
          "ask_cancellation_depth": 0.16078483285537223,
          "ask_cancellation_ticks": 0.16034055266294467,
          "ask_volume": 0.09541223691036332,
          "ask_volume_touch": 0.05560360135054905,
          "bid_cancellation_depth": 0.36614365151825323,
          "bid_cancellation_ticks": 0.3660282480702693,
          "bid_volume": 0.07380460650019885,
          "bid_volume_touch": 0.06149689618718125,
          "limit_ask_order_depth": 0.11817116504687425,
          "limit_ask_order_ticks": 0.11819818138644025,
          "limit_bid_order_depth": 0.36128147931077725,
          "limit_bid_order_ticks": 0.3605375514686384,
          "log_inter_arrival_time": 0.10949805827840052,
          "log_time_to_cancel": 0.23188787376605208,
          "ofi": 0.2983454605411607,
          "ofi_down": 0.269833749153752,
          "ofi_stay": 0.29991145969477195,
          "ofi_up": 0.23185605165781892,
          "orderbook_imbalance": 0.0899843723309798,
          "spread": 0.18953516001905793,
          "vol_per_min": 0.020096896259838903
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-1/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.19232152785569975,
            "report_record_id": "mag7-field10k-small-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-field10k-small-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-fixed26-large-s20000",
      "model": "Mamba3 fixed26 large seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.19041978897903153,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-fixed26-large-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.3160524661611581,
        "ask_cancellation_ticks": 0.3154151369975146,
        "ask_volume": 0.05186989269125381,
        "ask_volume_touch": 0.04038232406588674,
        "bid_cancellation_depth": 0.3078136753764601,
        "bid_cancellation_ticks": 0.30759485801447795,
        "bid_volume": 0.047759574305461495,
        "bid_volume_touch": 0.035474130832609,
        "limit_ask_order_depth": 0.24264508566471177,
        "limit_ask_order_ticks": 0.24259320629309788,
        "limit_bid_order_depth": 0.3198887807140642,
        "limit_bid_order_ticks": 0.318665050300454,
        "log_inter_arrival_time": 0.11546211654606181,
        "log_time_to_cancel": 0.4686953102950495,
        "ofi": 0.15783625486716313,
        "ofi_down": 0.20868063165892958,
        "ofi_stay": 0.15947789978091753,
        "ofi_up": 0.08360940276752443,
        "orderbook_imbalance": 0.08674597103153962,
        "spread": 0.13791527581961496,
        "vol_per_min": 0.03423852437571206
      },
      "evaluation_key": "68a94609816a220ce4fff9ed7e396c69ccb2499625fc8417d07095c2c6e1d8ce",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.19041978897903153,
      "g20_classification": {
        "id": "mag7-fixed26-large-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 fixed26 large seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.19041978897903153,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 995,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.19041978897903156,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-fixed26-large-s20000.json"
            }
          ],
          "evaluation_key": "68a94609816a220ce4fff9ed7e396c69ccb2499625fc8417d07095c2c6e1d8ce",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_large_aligned_seed42/attempts/attempt-0000/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_large_aligned_seed42/attempts/attempt-0000/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/generation/inference_manifest.json",
            "evidence_line": 404567,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_large_aligned_seed42/attempts/attempt-0000/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 404568,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "a243bce14693663e88e6c54e2d64671498b007d602db2db360a0c073a7c23a93",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_large_aligned_seed42/attempts/attempt-0000/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/generation/inference_manifest.json",
            "line": 404753,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/generation/inference_manifest.json",
            "line": 404755,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19041978897903156,
        "feature_points": {
          "ask_cancellation_depth": 0.3160524661611581,
          "ask_cancellation_ticks": 0.3154151369975146,
          "ask_volume": 0.05186989269125381,
          "ask_volume_touch": 0.04038232406588674,
          "bid_cancellation_depth": 0.3078136753764601,
          "bid_cancellation_ticks": 0.30759485801447795,
          "bid_volume": 0.047759574305461495,
          "bid_volume_touch": 0.035474130832609,
          "limit_ask_order_depth": 0.24264508566471177,
          "limit_ask_order_ticks": 0.24259320629309788,
          "limit_bid_order_depth": 0.3198887807140642,
          "limit_bid_order_ticks": 0.318665050300454,
          "log_inter_arrival_time": 0.11546211654606181,
          "log_time_to_cancel": 0.4686953102950495,
          "ofi": 0.15783625486716313,
          "ofi_down": 0.20868063165892958,
          "ofi_stay": 0.15947789978091753,
          "ofi_up": 0.08360940276752443,
          "orderbook_imbalance": 0.08674597103153962,
          "spread": 0.13791527581961496,
          "vol_per_min": 0.03423852437571206
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-8/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.19041978897903153,
            "report_record_id": "mag7-fixed26-large-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-fixed26-large-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-fixed26-medium-s20000",
      "model": "Mamba3 fixed26 medium seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.20648821274783935,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-fixed26-medium-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.38161843293588105,
        "ask_cancellation_ticks": 0.38134869334820487,
        "ask_volume": 0.03178454024615989,
        "ask_volume_touch": 0.028475117030261773,
        "bid_cancellation_depth": 0.35316688932283946,
        "bid_cancellation_ticks": 0.3531104142139456,
        "bid_volume": 0.021391263434013645,
        "bid_volume_touch": 0.03477123014664982,
        "limit_ask_order_depth": 0.27276686148029355,
        "limit_ask_order_ticks": 0.2727656732632395,
        "limit_bid_order_depth": 0.3704532958073184,
        "limit_bid_order_ticks": 0.36966980725458337,
        "log_inter_arrival_time": 0.13350131629823941,
        "log_time_to_cancel": 0.28508693543876296,
        "ofi": 0.2292343517211319,
        "ofi_down": 0.1595770138343906,
        "ofi_stay": 0.2315600006155183,
        "ofi_up": 0.17013667823709966,
        "orderbook_imbalance": 0.08991187413378618,
        "spread": 0.12250809813681944,
        "vol_per_min": 0.043413980805487216
      },
      "evaluation_key": "99039c086a6579de6eddfdf2261bb142a1813a26deac4446628d2ac7b5fc0cae",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.20648821274783935,
      "g20_classification": {
        "id": "mag7-fixed26-medium-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 fixed26 medium seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.20648821274783935,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 996,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.20648821274783938,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-fixed26-medium-s20000.json"
            }
          ],
          "evaluation_key": "99039c086a6579de6eddfdf2261bb142a1813a26deac4446628d2ac7b5fc0cae",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_medium_aligned_seed42/attempts/attempt-0001/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_medium_aligned_seed42/attempts/attempt-0001/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/generation/inference_manifest.json",
            "evidence_line": 404567,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_medium_aligned_seed42/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 404568,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "3d95a755b9e343c5248cbae51ef87d82178e4a719c8d3e01a50500b3470fc527",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_medium_aligned_seed42/attempts/attempt-0001/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/generation/inference_manifest.json",
            "line": 404753,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/generation/inference_manifest.json",
            "line": 404755,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.20648821274783938,
        "feature_points": {
          "ask_cancellation_depth": 0.38161843293588105,
          "ask_cancellation_ticks": 0.38134869334820487,
          "ask_volume": 0.03178454024615989,
          "ask_volume_touch": 0.028475117030261773,
          "bid_cancellation_depth": 0.35316688932283946,
          "bid_cancellation_ticks": 0.3531104142139456,
          "bid_volume": 0.021391263434013645,
          "bid_volume_touch": 0.03477123014664982,
          "limit_ask_order_depth": 0.27276686148029355,
          "limit_ask_order_ticks": 0.2727656732632395,
          "limit_bid_order_depth": 0.3704532958073184,
          "limit_bid_order_ticks": 0.36966980725458337,
          "log_inter_arrival_time": 0.13350131629823941,
          "log_time_to_cancel": 0.28508693543876296,
          "ofi": 0.2292343517211319,
          "ofi_down": 0.1595770138343906,
          "ofi_stay": 0.2315600006155183,
          "ofi_up": 0.17013667823709966,
          "orderbook_imbalance": 0.08991187413378618,
          "spread": 0.12250809813681944,
          "vol_per_min": 0.043413980805487216
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-4/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.20648821274783935,
            "report_record_id": "mag7-fixed26-medium-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-fixed26-medium-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-fixed26-small-s20000",
      "model": "Mamba3 fixed26 small seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.23058802830124212,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-fixed26-small-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.4029174979659441,
        "ask_cancellation_ticks": 0.402666983840788,
        "ask_volume": 0.08689330789842165,
        "ask_volume_touch": 0.06320343555020955,
        "bid_cancellation_depth": 0.3616662914867983,
        "bid_cancellation_ticks": 0.3614814958062813,
        "bid_volume": 0.046372853164643194,
        "bid_volume_touch": 0.0525901686586141,
        "limit_ask_order_depth": 0.27838112539544435,
        "limit_ask_order_ticks": 0.278466363955616,
        "limit_bid_order_depth": 0.3649339186162563,
        "limit_bid_order_ticks": 0.36419819770627304,
        "log_inter_arrival_time": 0.15525607347787523,
        "log_time_to_cancel": 0.37048301691507035,
        "ofi": 0.2638721684078149,
        "ofi_down": 0.1849724580752176,
        "ofi_stay": 0.2653132599166888,
        "ofi_up": 0.21127194972299723,
        "orderbook_imbalance": 0.08609989063353453,
        "spread": 0.1988959393252022,
        "vol_per_min": 0.0424121978063936
      },
      "evaluation_key": "e335365355c6ecbad1237f8ff452d783d3bf1a1679b65f06d36de44939e0c643",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.23058802830124212,
      "g20_classification": {
        "id": "mag7-fixed26-small-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 fixed26 small seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.23058802830124212,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 997,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.23058802830124212,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-fixed26-small-s20000.json"
            }
          ],
          "evaluation_key": "e335365355c6ecbad1237f8ff452d783d3bf1a1679b65f06d36de44939e0c643",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_small_aligned_seed42/attempts/attempt-0001/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_small_aligned_seed42/attempts/attempt-0001/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/generation/inference_manifest.json",
            "evidence_line": 404567,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_small_aligned_seed42/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 404568,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "5c77326805c68d2b5d59e38b65562a7509eb83a579d5b105f7c8c7c1142778a3",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_fixed26_small_aligned_seed42/attempts/attempt-0001/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/generation/inference_manifest.json",
            "line": 404753,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/generation/inference_manifest.json",
            "line": 404755,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.23058802830124212,
        "feature_points": {
          "ask_cancellation_depth": 0.4029174979659441,
          "ask_cancellation_ticks": 0.402666983840788,
          "ask_volume": 0.08689330789842165,
          "ask_volume_touch": 0.06320343555020955,
          "bid_cancellation_depth": 0.3616662914867983,
          "bid_cancellation_ticks": 0.3614814958062813,
          "bid_volume": 0.046372853164643194,
          "bid_volume_touch": 0.0525901686586141,
          "limit_ask_order_depth": 0.27838112539544435,
          "limit_ask_order_ticks": 0.278466363955616,
          "limit_bid_order_depth": 0.3649339186162563,
          "limit_bid_order_ticks": 0.36419819770627304,
          "log_inter_arrival_time": 0.15525607347787523,
          "log_time_to_cancel": 0.37048301691507035,
          "ofi": 0.2638721684078149,
          "ofi_down": 0.1849724580752176,
          "ofi_stay": 0.2653132599166888,
          "ofi_up": 0.21127194972299723,
          "orderbook_imbalance": 0.08609989063353453,
          "spread": 0.1988959393252022,
          "vol_per_min": 0.0424121978063936
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-0/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.23058802830124212,
            "report_record_id": "mag7-fixed26-small-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-fixed26-small-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-message15k-large-s15000",
      "model": "Mamba3 message15k large seed 42",
      "evaluation": "step 15000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.16303191711851606,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message15k-large-s15000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.16081772341591527,
        "ask_cancellation_ticks": 0.1604186432294059,
        "ask_volume": 0.06644620606819772,
        "ask_volume_touch": 0.06731105484319994,
        "bid_cancellation_depth": 0.31920449783160254,
        "bid_cancellation_ticks": 0.3193336914475265,
        "bid_volume": 0.049465132534398965,
        "bid_volume_touch": 0.06673496997139436,
        "limit_ask_order_depth": 0.12557485986988365,
        "limit_ask_order_ticks": 0.12565069668678455,
        "limit_bid_order_depth": 0.343724666711511,
        "limit_bid_order_ticks": 0.342767290134991,
        "log_inter_arrival_time": 0.05764302737406825,
        "log_time_to_cancel": 0.23119300216468244,
        "ofi": 0.16459858250750858,
        "ofi_down": 0.14100727025854326,
        "ofi_stay": 0.16652036664756248,
        "ofi_up": 0.09280178489000843,
        "orderbook_imbalance": 0.1024081340870249,
        "spread": 0.2723848450321254,
        "vol_per_min": 0.04766381378250182
      },
      "evaluation_key": "bfbc316d0f632fcd711efef54ac3730e2541876fcc1d100b058264de0b9b0447",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.16303191711851606,
      "g20_classification": {
        "id": "mag7-message15k-large-s15000",
        "class": "G20",
        "model_evaluation": "Mamba3 message15k large seed 42 | step 15000; pooled January validation",
        "wws21_mean": 0.16303191711851606,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 998,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.16303191711851606,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message15k-large-s15000.json"
            }
          ],
          "evaluation_key": "bfbc316d0f632fcd711efef54ac3730e2541876fcc1d100b058264de0b9b0447",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_large_aligned_seed42/attempts/attempt-0000/run/15000",
            "path_kind": "directory",
            "checkpoint_step": 15000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_large_aligned_seed42/attempts/attempt-0000/run/15000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_large_aligned_seed42/attempts/attempt-0000/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 15000,",
            "score_manifest_path": "/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "e2f0f4ad9531c7e4e78eafd176ef355234032e0bec6b8e9ffa00693ce0d7046b",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_large_aligned_seed42/attempts/attempt-0000/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.16303191711851606,
        "feature_points": {
          "ask_cancellation_depth": 0.16081772341591527,
          "ask_cancellation_ticks": 0.1604186432294059,
          "ask_volume": 0.06644620606819772,
          "ask_volume_touch": 0.06731105484319994,
          "bid_cancellation_depth": 0.31920449783160254,
          "bid_cancellation_ticks": 0.3193336914475265,
          "bid_volume": 0.049465132534398965,
          "bid_volume_touch": 0.06673496997139436,
          "limit_ask_order_depth": 0.12557485986988365,
          "limit_ask_order_ticks": 0.12565069668678455,
          "limit_bid_order_depth": 0.343724666711511,
          "limit_bid_order_ticks": 0.342767290134991,
          "log_inter_arrival_time": 0.05764302737406825,
          "log_time_to_cancel": 0.23119300216468244,
          "ofi": 0.16459858250750858,
          "ofi_down": 0.14100727025854326,
          "ofi_stay": 0.16652036664756248,
          "ofi_up": 0.09280178489000843,
          "orderbook_imbalance": 0.1024081340870249,
          "spread": 0.2723848450321254,
          "vol_per_min": 0.04766381378250182
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/experiments/mag7-message15k-large-step15000/job-6429362/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.16303191711851606,
            "report_record_id": "mag7-message15k-large-s15000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-message15k-large-s15000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-message15k-large-s20000",
      "model": "Mamba3 message15k large seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.19090937952818599,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message15k-large-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.1291370407199428,
        "ask_cancellation_ticks": 0.12899478912513748,
        "ask_volume": 0.10630133535924124,
        "ask_volume_touch": 0.07332483699215597,
        "bid_cancellation_depth": 0.32027775641743006,
        "bid_cancellation_ticks": 0.3205397904127495,
        "bid_volume": 0.08487405237955763,
        "bid_volume_touch": 0.06643173319115411,
        "limit_ask_order_depth": 0.1181996358962018,
        "limit_ask_order_ticks": 0.11822614536627676,
        "limit_bid_order_depth": 0.33336076716814006,
        "limit_bid_order_ticks": 0.3323579572571098,
        "log_inter_arrival_time": 0.13403606487370798,
        "log_time_to_cancel": 0.44640554316911873,
        "ofi": 0.296263650608208,
        "ofi_down": 0.18313535067995398,
        "ofi_stay": 0.29801131969175176,
        "ofi_up": 0.19412728934689513,
        "orderbook_imbalance": 0.09140163789665309,
        "spread": 0.19275873228655344,
        "vol_per_min": 0.04093154125396662
      },
      "evaluation_key": "f7d34a636e3debe08bc4fe4c00443095a7f60ec87aec93eede95c22edb02c4f5",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.19090937952818599,
      "g20_classification": {
        "id": "mag7-message15k-large-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 message15k large seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.19090937952818599,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 999,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.19090937952818599,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message15k-large-s20000.json"
            }
          ],
          "evaluation_key": "f7d34a636e3debe08bc4fe4c00443095a7f60ec87aec93eede95c22edb02c4f5",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_large_aligned_seed42/attempts/attempt-0000/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_large_aligned_seed42/attempts/attempt-0000/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_large_aligned_seed42/attempts/attempt-0000/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "f0b63456a424f1ed2916c568791d1c700f7aad813c0adc8f8005f8ed8825ad05",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_large_aligned_seed42/attempts/attempt-0000/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19090937952818599,
        "feature_points": {
          "ask_cancellation_depth": 0.1291370407199428,
          "ask_cancellation_ticks": 0.12899478912513748,
          "ask_volume": 0.10630133535924124,
          "ask_volume_touch": 0.07332483699215597,
          "bid_cancellation_depth": 0.32027775641743006,
          "bid_cancellation_ticks": 0.3205397904127495,
          "bid_volume": 0.08487405237955763,
          "bid_volume_touch": 0.06643173319115411,
          "limit_ask_order_depth": 0.1181996358962018,
          "limit_ask_order_ticks": 0.11822614536627676,
          "limit_bid_order_depth": 0.33336076716814006,
          "limit_bid_order_ticks": 0.3323579572571098,
          "log_inter_arrival_time": 0.13403606487370798,
          "log_time_to_cancel": 0.44640554316911873,
          "ofi": 0.296263650608208,
          "ofi_down": 0.18313535067995398,
          "ofi_stay": 0.29801131969175176,
          "ofi_up": 0.19412728934689513,
          "orderbook_imbalance": 0.09140163789665309,
          "spread": 0.19275873228655344,
          "vol_per_min": 0.04093154125396662
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-11/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.19090937952818599,
            "report_record_id": "mag7-message15k-large-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-message15k-large-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-message15k-medium-s20000",
      "model": "Mamba3 message15k medium seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.1666523839445424,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message15k-medium-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.1267566542069773,
        "ask_cancellation_ticks": 0.1264460718976399,
        "ask_volume": 0.11654834357485674,
        "ask_volume_touch": 0.07059110993349242,
        "bid_cancellation_depth": 0.2951441183303729,
        "bid_cancellation_ticks": 0.29530608838181993,
        "bid_volume": 0.09740459087959116,
        "bid_volume_touch": 0.08185265048684211,
        "limit_ask_order_depth": 0.15425613510058547,
        "limit_ask_order_ticks": 0.15407739893343167,
        "limit_bid_order_depth": 0.31592812934105213,
        "limit_bid_order_ticks": 0.3150378215606764,
        "log_inter_arrival_time": 0.027425275270726315,
        "log_time_to_cancel": 0.1965614822998805,
        "ofi": 0.18797296703021035,
        "ofi_down": 0.20607533442582668,
        "ofi_stay": 0.188661573734047,
        "ofi_up": 0.13371877648023095,
        "orderbook_imbalance": 0.0969300199895409,
        "spread": 0.27201359340314313,
        "vol_per_min": 0.0409919275744465
      },
      "evaluation_key": "9907930c29c4bc021cf29082583661808c691d0bd54a67e2472d1d66870b90fe",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.1666523839445424,
      "g20_classification": {
        "id": "mag7-message15k-medium-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 message15k medium seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.1666523839445424,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 1000,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.16665238394454243,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message15k-medium-s20000.json"
            }
          ],
          "evaluation_key": "9907930c29c4bc021cf29082583661808c691d0bd54a67e2472d1d66870b90fe",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_medium_aligned_seed42/attempts/attempt-0001/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_medium_aligned_seed42/attempts/attempt-0001/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_medium_aligned_seed42/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "ec237db4f9e823399a83a814b1f668350243e3c904d988b1b7dcf78f6a81d76c",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_medium_aligned_seed42/attempts/attempt-0001/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.1666523839445424,
        "feature_points": {
          "ask_cancellation_depth": 0.1267566542069773,
          "ask_cancellation_ticks": 0.1264460718976399,
          "ask_volume": 0.11654834357485674,
          "ask_volume_touch": 0.07059110993349242,
          "bid_cancellation_depth": 0.2951441183303729,
          "bid_cancellation_ticks": 0.29530608838181993,
          "bid_volume": 0.09740459087959116,
          "bid_volume_touch": 0.08185265048684211,
          "limit_ask_order_depth": 0.15425613510058547,
          "limit_ask_order_ticks": 0.15407739893343167,
          "limit_bid_order_depth": 0.31592812934105213,
          "limit_bid_order_ticks": 0.3150378215606764,
          "log_inter_arrival_time": 0.027425275270726315,
          "log_time_to_cancel": 0.1965614822998805,
          "ofi": 0.18797296703021035,
          "ofi_down": 0.20607533442582668,
          "ofi_stay": 0.188661573734047,
          "ofi_up": 0.13371877648023095,
          "orderbook_imbalance": 0.0969300199895409,
          "spread": 0.27201359340314313,
          "vol_per_min": 0.0409919275744465
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-7/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.1666523839445424,
            "report_record_id": "mag7-message15k-medium-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-message15k-medium-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-message15k-small-s20000",
      "model": "Mamba3 message15k small seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.17946593219491788,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message15k-small-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.09869618744416538,
        "ask_cancellation_ticks": 0.09848022896427167,
        "ask_volume": 0.12545963715419955,
        "ask_volume_touch": 0.09303670150800752,
        "bid_cancellation_depth": 0.3297858472903188,
        "bid_cancellation_ticks": 0.32993385159941774,
        "bid_volume": 0.0928773232533413,
        "bid_volume_touch": 0.07858456563989842,
        "limit_ask_order_depth": 0.20120421664248314,
        "limit_ask_order_ticks": 0.2012889077191985,
        "limit_bid_order_depth": 0.34043833854521977,
        "limit_bid_order_ticks": 0.3396315070877627,
        "log_inter_arrival_time": 0.03973460435690026,
        "log_time_to_cancel": 0.2069794170367673,
        "ofi": 0.18372998935637816,
        "ofi_down": 0.11842398600495588,
        "ofi_stay": 0.18455035973665995,
        "ofi_up": 0.2230664975054116,
        "orderbook_imbalance": 0.10650305202550833,
        "spread": 0.3286214327523747,
        "vol_per_min": 0.047757924470034636
      },
      "evaluation_key": "6ebd2a1466b76f45fb7697c96262d10c89a989b639d3a00f243135162959277e",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.17946593219491788,
      "g20_classification": {
        "id": "mag7-message15k-small-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 message15k small seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.17946593219491788,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 1001,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.17946593219491788,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message15k-small-s20000.json"
            }
          ],
          "evaluation_key": "6ebd2a1466b76f45fb7697c96262d10c89a989b639d3a00f243135162959277e",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_small_aligned_seed42/attempts/attempt-0002/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_small_aligned_seed42/attempts/attempt-0002/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_small_aligned_seed42/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "0c465729cf428c9b38cd234359ff284d8dee50edd9b4033234c5edc2360c17a1",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message15k_small_aligned_seed42/attempts/attempt-0002/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.17946593219491788,
        "feature_points": {
          "ask_cancellation_depth": 0.09869618744416538,
          "ask_cancellation_ticks": 0.09848022896427167,
          "ask_volume": 0.12545963715419955,
          "ask_volume_touch": 0.09303670150800752,
          "bid_cancellation_depth": 0.3297858472903188,
          "bid_cancellation_ticks": 0.32993385159941774,
          "bid_volume": 0.0928773232533413,
          "bid_volume_touch": 0.07858456563989842,
          "limit_ask_order_depth": 0.20120421664248314,
          "limit_ask_order_ticks": 0.2012889077191985,
          "limit_bid_order_depth": 0.34043833854521977,
          "limit_bid_order_ticks": 0.3396315070877627,
          "log_inter_arrival_time": 0.03973460435690026,
          "log_time_to_cancel": 0.2069794170367673,
          "ofi": 0.18372998935637816,
          "ofi_down": 0.11842398600495588,
          "ofi_stay": 0.18455035973665995,
          "ofi_up": 0.2230664975054116,
          "orderbook_imbalance": 0.10650305202550833,
          "spread": 0.3286214327523747,
          "vol_per_min": 0.047757924470034636
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-3/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.17946593219491788,
            "report_record_id": "mag7-message15k-small-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-message15k-small-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-message5k-large-s20000",
      "model": "Mamba3 message5k large seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.16939676590261432,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message5k-large-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.16732584497882824,
        "ask_cancellation_ticks": 0.16683719863418445,
        "ask_volume": 0.09813982469590155,
        "ask_volume_touch": 0.06955589041694987,
        "bid_cancellation_depth": 0.3209153125989346,
        "bid_cancellation_ticks": 0.3199508894522638,
        "bid_volume": 0.03945414426615848,
        "bid_volume_touch": 0.05390623949552839,
        "limit_ask_order_depth": 0.181198858826135,
        "limit_ask_order_ticks": 0.18095927321028213,
        "limit_bid_order_depth": 0.3142212651120263,
        "limit_bid_order_ticks": 0.31302573414113205,
        "log_inter_arrival_time": 0.027683388777735088,
        "log_time_to_cancel": 0.2418000900637743,
        "ofi": 0.175809837407076,
        "ofi_down": 0.12911967265003108,
        "ofi_stay": 0.17695329508780128,
        "ofi_up": 0.18396528379320182,
        "orderbook_imbalance": 0.10707422752627896,
        "spread": 0.247981424727693,
        "vol_per_min": 0.04145438809298418
      },
      "evaluation_key": "4207b3c7e926712d55f926769fbf4cfdac822f048ad71c53676345dba27f7a9f",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.16939676590261432,
      "g20_classification": {
        "id": "mag7-message5k-large-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 message5k large seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.16939676590261432,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 1002,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.16939676590261432,
          "absolute_difference": 0.0,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message5k-large-s20000.json"
            }
          ],
          "evaluation_key": "4207b3c7e926712d55f926769fbf4cfdac822f048ad71c53676345dba27f7a9f",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/job-6401855/checkpoints/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/job-6401855/checkpoints/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/job-6401855/checkpoints\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "6b58197599ee69a46d0d32f012e834ffa2705b9015e4135c4fa87f16b7a8f747",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/job-6401855/checkpoints"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.16939676590261432,
        "feature_points": {
          "ask_cancellation_depth": 0.16732584497882824,
          "ask_cancellation_ticks": 0.16683719863418445,
          "ask_volume": 0.09813982469590155,
          "ask_volume_touch": 0.06955589041694987,
          "bid_cancellation_depth": 0.3209153125989346,
          "bid_cancellation_ticks": 0.3199508894522638,
          "bid_volume": 0.03945414426615848,
          "bid_volume_touch": 0.05390623949552839,
          "limit_ask_order_depth": 0.181198858826135,
          "limit_ask_order_ticks": 0.18095927321028213,
          "limit_bid_order_depth": 0.3142212651120263,
          "limit_bid_order_ticks": 0.31302573414113205,
          "log_inter_arrival_time": 0.027683388777735088,
          "log_time_to_cancel": 0.2418000900637743,
          "ofi": 0.175809837407076,
          "ofi_down": 0.12911967265003108,
          "ofi_stay": 0.17695329508780128,
          "ofi_up": 0.18396528379320182,
          "orderbook_imbalance": 0.10707422752627896,
          "spread": 0.247981424727693,
          "vol_per_min": 0.04145438809298418
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/experiments/mag7-message5k-large-muon005-recovery/lobbench/job-6427396/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.16939676590261432,
            "report_record_id": "mag7-message5k-large-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-message5k-large-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-message5k-medium-s20000",
      "model": "Mamba3 message5k medium seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.184036201930208,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message5k-medium-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.2648194181467435,
        "ask_cancellation_ticks": 0.26448597502265664,
        "ask_volume": 0.08504273755725857,
        "ask_volume_touch": 0.07312550865575718,
        "bid_cancellation_depth": 0.32247030985664177,
        "bid_cancellation_ticks": 0.3215641053956979,
        "bid_volume": 0.12404060960607069,
        "bid_volume_touch": 0.08501665827682751,
        "limit_ask_order_depth": 0.16337673090302024,
        "limit_ask_order_ticks": 0.16328088356478285,
        "limit_bid_order_depth": 0.3358752418469637,
        "limit_bid_order_ticks": 0.33547850229509735,
        "log_inter_arrival_time": 0.05195253118215135,
        "log_time_to_cancel": 0.34593318620660135,
        "ofi": 0.09676675207425986,
        "ofi_down": 0.15770188994411527,
        "ofi_stay": 0.09698933078647036,
        "ofi_up": 0.05354113909792469,
        "orderbook_imbalance": 0.10081514404941322,
        "spread": 0.37524766555855377,
        "vol_per_min": 0.04723592050736058
      },
      "evaluation_key": "8bac42d352fec2d5979c172ecc212b5c8467e38d1f4f8439ef96ed4859a2c4d7",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.184036201930208,
      "g20_classification": {
        "id": "mag7-message5k-medium-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 message5k medium seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.184036201930208,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 1003,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.18403620193020803,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message5k-medium-s20000.json"
            }
          ],
          "evaluation_key": "8bac42d352fec2d5979c172ecc212b5c8467e38d1f4f8439ef96ed4859a2c4d7",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message5k_medium_aligned_seed42/attempts/attempt-0001/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message5k_medium_aligned_seed42/attempts/attempt-0001/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message5k_medium_aligned_seed42/attempts/attempt-0001/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "2b3a5526550a663876736bf8e681217a25cc53e6acd5c11cf67a332b7803e652",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message5k_medium_aligned_seed42/attempts/attempt-0001/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.18403620193020803,
        "feature_points": {
          "ask_cancellation_depth": 0.2648194181467435,
          "ask_cancellation_ticks": 0.26448597502265664,
          "ask_volume": 0.08504273755725857,
          "ask_volume_touch": 0.07312550865575718,
          "bid_cancellation_depth": 0.32247030985664177,
          "bid_cancellation_ticks": 0.3215641053956979,
          "bid_volume": 0.12404060960607069,
          "bid_volume_touch": 0.08501665827682751,
          "limit_ask_order_depth": 0.16337673090302024,
          "limit_ask_order_ticks": 0.16328088356478285,
          "limit_bid_order_depth": 0.3358752418469637,
          "limit_bid_order_ticks": 0.33547850229509735,
          "log_inter_arrival_time": 0.05195253118215135,
          "log_time_to_cancel": 0.34593318620660135,
          "ofi": 0.09676675207425986,
          "ofi_down": 0.15770188994411527,
          "ofi_stay": 0.09698933078647036,
          "ofi_up": 0.05354113909792469,
          "orderbook_imbalance": 0.10081514404941322,
          "spread": 0.37524766555855377,
          "vol_per_min": 0.04723592050736058
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-6/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.184036201930208,
            "report_record_id": "mag7-message5k-medium-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-message5k-medium-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    },
    {
      "id": "mag7-message5k-small-s20000",
      "model": "Mamba3 message5k small seed 42",
      "evaluation": "step 20000; pooled January validation",
      "protocol_id": "mag7-jan26",
      "mean_ws21": 0.19102817828343843,
      "evaluated_at": null,
      "date_evidence": null,
      "evidence": [
        {
          "label": "Feature values & provenance",
          "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message5k-small-s20000.json"
        }
      ],
      "features": {
        "ask_cancellation_depth": 0.18007924631407096,
        "ask_cancellation_ticks": 0.17981459499667204,
        "ask_volume": 0.07935010475206386,
        "ask_volume_touch": 0.07795785484004683,
        "bid_cancellation_depth": 0.3490813677978124,
        "bid_cancellation_ticks": 0.3480152972536296,
        "bid_volume": 0.07896239535416924,
        "bid_volume_touch": 0.07701421091308225,
        "limit_ask_order_depth": 0.19840764816779677,
        "limit_ask_order_ticks": 0.19786604307132064,
        "limit_bid_order_depth": 0.35447914193121466,
        "limit_bid_order_ticks": 0.35411989071530436,
        "log_inter_arrival_time": 0.07877576911155741,
        "log_time_to_cancel": 0.31368665178462,
        "ofi": 0.11138052853981616,
        "ofi_down": 0.22161796215394935,
        "ofi_stay": 0.10926205899990922,
        "ofi_up": 0.06507392045117746,
        "orderbook_imbalance": 0.10133920776007718,
        "spread": 0.5008502269383519,
        "vol_per_min": 0.034457622105565064
      },
      "evaluation_key": "40f3bd8f0183348ddb50e3bc5f4e23850157ff1e5a3d12a97e1cbe91284bf9e6",
      "date": "2026-09-09",
      "collection_source": "existing_verified_results",
      "date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "date_is_evaluation_completion": false,
      "archive_source_date": "2026-09-09",
      "archive_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "source_date": "2026-09-09",
      "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
      "evaluation_completed_at": null,
      "class": "G20",
      "wws21_mean": 0.19102817828343843,
      "g20_classification": {
        "id": "mag7-message5k-small-s20000",
        "class": "G20",
        "model_evaluation": "Mamba3 message5k small seed 42 | step 20000; pooled January validation",
        "wws21_mean": 0.19102817828343843,
        "stock": "GOOG",
        "market_dates": [
          "2026-01-02",
          "2026-01-05",
          "2026-01-06",
          "2026-01-07",
          "2026-01-08",
          "2026-01-09",
          "2026-01-12",
          "2026-01-13",
          "2026-01-14",
          "2026-01-15",
          "2026-01-16",
          "2026-01-20",
          "2026-01-21",
          "2026-01-22",
          "2026-01-23",
          "2026-01-26",
          "2026-01-27",
          "2026-01-28",
          "2026-01-29",
          "2026-01-30"
        ],
        "market_trading_day_count": 20,
        "market_date_status": "verified",
        "date_confidence": "verified",
        "recorded_date": null,
        "recorded_date_status": "unknown",
        "recorded_date_evidence": null,
        "source_recorded_date": "2026-09-09",
        "source_date_basis": "registry metadata date_experiment: mag7 scaling experiment",
        "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/scores/score_manifest.json",
        "input_path": "",
        "path_type": "score_summary",
        "path_status": "exists",
        "notes": "GOOG January 2026；覆盖 20 个交易日；评估完成日期未确认；来源原有日期 2026-09-09（registry metadata date_experiment: mag7 scaling experiment）；不作为评估完成时间；path 类型：score_summary；protocol=mag7-jan26",
        "protocol": "mag7-jan26",
        "score_status": "21_feature_mean_verified",
        "source_url": null,
        "source_selector": "pooled/wasserstein_by_feature",
        "snapshot_revision": "3ed167f2c59fc576b0081513aa35d89313c52b83",
        "source_table_row_index": 1004,
        "full21_provenance": {
          "feature_count": 21,
          "validation": "arithmetic mean independently rechecked against original public per-feature values",
          "recomputed_mean": 0.19102817828343846,
          "absolute_difference": 2.7755575615628914e-17,
          "evidence": [
            {
              "label": "Feature values & provenance",
              "url": "https://huggingface.co/spaces/oxford-lob/leaderboard/blob/main/data/evidence/mag7-message5k-small-s20000.json"
            }
          ],
          "evaluation_key": "40f3bd8f0183348ddb50e3bc5f4e23850157ff1e5a3d12a97e1cbe91284bf9e6",
          "source_artifact": null,
          "source_sha256": null
        }
      },
      "checkpoint_info": {
        "status": "found",
        "checkpoints": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message5k_small_aligned_seed42/attempts/attempt-0002/run/20000",
            "path_kind": "directory",
            "checkpoint_step": 20000,
            "model_seed": 42,
            "availability": "exists",
            "selected_step_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message5k_small_aligned_seed42/attempts/attempt-0002/run/20000",
            "selected_step_availability": "exists",
            "availability_scope": "selected_checkpoint_step_directory",
            "evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/generation/inference_manifest.json",
            "evidence_line": 417111,
            "evidence_text": "\"root\": \"/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message5k_small_aligned_seed42/attempts/attempt-0002/run\",",
            "checkpoint_step_evidence_path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/generation/inference_manifest.json",
            "checkpoint_step_evidence_line": 417112,
            "checkpoint_step_evidence_text": "\"step\": 20000,",
            "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/scores/score_manifest.json",
            "recorded_checkpoint_tree_sha256": "f76752c0b5ab08b2fd78fd73600ae80fd6ba05225058ab4d50c4a2b0bf5dc33a",
            "recorded_checkpoint_root": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/mag7_2025_aligned_scaling_v1/checkpoints/mag7_2025_message5k_small_aligned_seed42/attempts/attempt-0002/run"
          }
        ],
        "note": "The score manifest explicitly references this inference manifest, which records the checkpoint root and selected step. The selected step directory exists; weights were not read.",
        "completeness": "complete",
        "expected_component_count": 1,
        "resolved_component_count": 1
      },
      "length_info": {
        "context_length": 250,
        "generation_length": 250,
        "unit": "messages",
        "status": "found",
        "evidence": [
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/generation/inference_manifest.json",
            "line": 417297,
            "text": "\"conditioning_messages\": 250,"
          },
          {
            "path": "/lus/lfs1aip2/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/generation/inference_manifest.json",
            "line": 417299,
            "text": "\"generated_messages\": 250,"
          }
        ],
        "note": "The run inference manifest explicitly records C250/G250. Its metadata inventory independently confirms 250 conditioning-message rows and 250 generated-message rows in each of all 3136 windows. Book snapshot rows and generated token counts are separate quantities.",
        "evidence_scope": "Run-specific inference protocol and all recorded per-window metadata inventories.",
        "verified_window_count": 3136,
        "score_manifest_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/scores/score_manifest.json"
      },
      "score_audit_info": {
        "status": "verified_21",
        "original_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/scores/score_manifest.json",
        "display_path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/scores/score_manifest.json",
        "path_resolution": "original_recorded_path",
        "feature_count": 21,
        "finite_feature_count": 21,
        "nonfinite_features": [],
        "recomputed_mean": 0.19102817828343846,
        "feature_points": {
          "ask_cancellation_depth": 0.18007924631407096,
          "ask_cancellation_ticks": 0.17981459499667204,
          "ask_volume": 0.07935010475206386,
          "ask_volume_touch": 0.07795785484004683,
          "bid_cancellation_depth": 0.3490813677978124,
          "bid_cancellation_ticks": 0.3480152972536296,
          "bid_volume": 0.07896239535416924,
          "bid_volume_touch": 0.07701421091308225,
          "limit_ask_order_depth": 0.19840764816779677,
          "limit_ask_order_ticks": 0.19786604307132064,
          "limit_bid_order_depth": 0.35447914193121466,
          "limit_bid_order_ticks": 0.35411989071530436,
          "log_inter_arrival_time": 0.07877576911155741,
          "log_time_to_cancel": 0.31368665178462,
          "ofi": 0.11138052853981616,
          "ofi_down": 0.22161796215394935,
          "ofi_stay": 0.10926205899990922,
          "ofi_up": 0.06507392045117746,
          "orderbook_imbalance": 0.10133920776007718,
          "spread": 0.5008502269383519,
          "vol_per_min": 0.034457622105565064
        },
        "evidence": [
          {
            "kind": "saved_public_feature_vector",
            "path": "/projects/public/u6gb/sigma-0/artifacts/mamba3_variable_bpe/january_lobbench_mag7_2025_scaling_goog_v2/task-2/scores/score_manifest.json",
            "selector": "pooled/wasserstein_by_feature",
            "source_value": 0.19102817828343843,
            "report_record_id": "mag7-message5k-small-s20000",
            "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
            "feature_vector_url": "https://huggingface.co/spaces/oxford-lob/leaderboard/resolve/9da61c67a4842026a15013a1170ceec303b48572/data/g20-results.json",
            "feature_vector_selector": "results[id=mag7-message5k-small-s20000].features"
          }
        ],
        "note": "Arithmetic mean independently recomputed from 21 saved finite feature values and matched to the reported source value within its recorded precision. Model scoring was not rerun."
      }
    }
  ],
  "checkpoint_audit": {
    "source_revision": "bbfcc16ba483c43af83cf82ece117695468dc12d",
    "checked_at": "2026-09-12T06:19:33.306318+00:00",
    "status_counts": {
      "found": 347,
      "not_found": 10
    },
    "limitations": [
      "Bounded audit of the 357 published dated G20 records, using evaluation logs, manifests, recorded input reuse, and shallow checkpoint directory metadata.",
      "No weights were opened, hashed, restored or used for inference. Directory existence does not prove integrity or loadability.",
      "Historical-to-public prefix mappings were only accepted when the exact run and requested step directory were verified.",
      "Copies in _DELETING_ directories remain unconfirmed relocation candidates; their existence does not change the primary checkpoint availability.",
      "Eight multi-seed aggregate rows resolve only seeds 5 and 137; seed 42 is unresolved because its source metadata is inaccessible.",
      "Not found means no supported checkpoint mapping was recovered within the audited metadata; it does not prove no checkpoint exists elsewhere."
    ]
  },
  "length_audit": {
    "source_revision": "9da61c67a4842026a15013a1170ceec303b48572",
    "checked_at": "2026-09-12T06:49:03.622749+00:00",
    "definitions": {
      "context_length": "Configured number of conditioning LOB messages supplied for this inference/evaluation, not training token context.",
      "generation_length": "Configured number of LOB messages generated per inference sequence, excluding conditioning messages; distinct from any later scoring truncation."
    },
    "status_counts": {
      "found": 357
    }
  },
  "score_audit": {
    "counts": {
      "verified_21": 121,
      "incomplete_21": 6,
      "source_only": 230
    },
    "source_report_sha256": "e78c863251e691c459b53e05502cf3feb403c1eea81facd376f87f1cda8b70e8",
    "checked_at": "2026-09-12T07:14:07.697356+00:00"
  }
}