{
  "study": "FinQA published test-split evidence retrieval",
  "executed_at": "2026-11-01T23:15:30.375849+00:00",
  "evidence_class": "real_public_financial_documents",
  "task": "rank supplied text sentences and table rows; not answer generation or OCR",
  "configuration": {
    "bm25_k1": 1.5,
    "bm25_b": 0.75,
    "k": [
      1,
      3,
      5
    ],
    "fitting": "document-local unsupervised vocabulary only",
    "test_tuning": "none; one configuration per baseline"
  },
  "examples": 1147,
  "documents": 380,
  "source_sha256": "831dbfb2e785dbc227f895ce3f24046433467aec67b09db2bd6ac7692a8a30dc",
  "script_sha256": "175fce1cf63c3a081ce276fca3b9351b712b21110ff66c2f327163c30ef509f7",
  "results": {
    "bm25": {
      "1": {
        "recall": 0.3911145015983726,
        "complete": 0.2833478639930253
      },
      "3": {
        "recall": 0.6245568148793955,
        "complete": 0.4925893635571055
      },
      "5": {
        "recall": 0.7390366172624236,
        "complete": 0.6181342632955537
      }
    },
    "tfidf": {
      "1": {
        "recall": 0.33934902644580067,
        "complete": 0.24585876198779424
      },
      "3": {
        "recall": 0.5617843650101714,
        "complete": 0.43591979075850046
      },
      "5": {
        "recall": 0.6959677419354838,
        "complete": 0.5789014821272885
      }
    }
  },
  "frontier_runs": [],
  "amlok_foundation_model_runs": [],
  "paired_comparison": {
    "metric": "recall_at_5",
    "difference": "bm25 minus tfidf",
    "point": 0.04306887532693985,
    "ci95": [
      0.023691578364463603,
      0.06221036540948372
    ],
    "resampling_unit": "source filename",
    "replicates": 2000,
    "seed": 20261101,
    "scope": "conditional on this corpus; not independent institutions or Canadian borrowers"
  },
  "predictions_sha256": "7b375d86e2ea6f87e95c8b5cad1493700c723034f03ec6387b169b0d3641f625"
}