{
  "status": "VERIFIED_COMPLETE_THREE_1342",
  "question_count": 1342,
  "added_model_count": 3,
  "primary_attempts": 4026,
  "recoveries": 0,
  "unusable_responses": 0,
  "bank_sha256": "2db8cf8353f12b609d4df5330d94f53390c9c2919677dc793df4bb2d3cff7076",
  "source_summary_sha256": "76beb2eeb9e44991e7958eb0dcbaa1bfa645e23ba6d780ed4ee652a73aad563f",
  "evidence_sha256": {
    "mammoth1": {
      "attempts.jsonl": "85d26db189ce9096a1aad886bf5a43c8759e9516d718ee05fd72ade68b0113da",
      "compatibility_probe.json": "40ffbb0ff792c797b17de0e7d5bdee1ea7bf81e84ac85d4e95edbb3d8c3c61bd",
      "completion.json": "dd2ace518c16a82cee76ad4a46a9aab5aac9d876c62f7593d3d9cae3792a6bd5",
      "dispatches.jsonl": "f4872c3590b70d7b4fab209e9c7d398f077391d3625110503f58abd4f7ff8b47",
      "manifest.json": "ffb87f00523800daec36cbaad4f37ca6967f95b71c8b0b31cba0df3d7d78fa9e",
      "model_metadata.json": "5cc4938d86520cb95a30f1efea78712e754903b7664567a7f37dae8543044976",
      "probe_dispatch.json": "75c66b4c40f903244e9df02b25541009da7654ace5ea6492dc86212abeb8bd99",
      "results.jsonl": "5fd083f6b57b2f22fdeec9502ea5e0d79d82461229f1651fe29dd7a72e75e369",
      "software.json": "05b7e4270efa627a24b7c94c30526887efb3a71408dc4166074f68dc13716655"
    },
    "elephant1": {
      "attempts.jsonl": "e18e2b43cf3e82c10e4abdac964b6ef2c5baba083ac68e3e53f832da8245e982",
      "compatibility_probe.json": "8ec896463b98ae132f749daf8beb7843eaa2f277660f41909a16ff73100b74a3",
      "completion.json": "3ace9a121889bfe4cf05e8be04961ca9bf7f99ade76f5c9dfa0974fe17134ea6",
      "dispatches.jsonl": "9b943e0585a50f2287d535a513650fac1e5134d0d2282f236bcb7ddd33ecb57d",
      "manifest.json": "d37530439b32ac0f20948db2396652198a4eb6d95d1e3beaa3410b8d36aff0dc",
      "model_metadata.json": "083b4cec56c7709d76b28b8ceca6f3acac49061e9e8ad0b2effd7b82f7b53550",
      "probe_dispatch.json": "13eeeb53ccf15a2a998e1f4a0121b7e7213e8a8d9ac99831d3868c39116ec0e1",
      "results.jsonl": "cbfda4df0cbde86b21c1dc36d62938e91558eaeef6404f48ff8fa6d207f6a412",
      "software.json": "05b7e4270efa627a24b7c94c30526887efb3a71408dc4166074f68dc13716655"
    },
    "drex_max": {
      "admission_note.json": "2d96b12d8e62787b94c81a0cca0fd6a202d28a10c54db070999f410d02283adc",
      "attempts.jsonl": "165a7f64bb7af7f27bb03f5f1fb467c46c8595f68da9cf4f5d37c7aca4da381c",
      "compatibility_probe.json": "fccd9a0beaeb376a73f121f7583f9088da24c0bacfe8ea1111a2f0f120182194",
      "completion.json": "b78fa8fcbf54d26e9365e1d7df893066c47165a1e543f050d810e881362bc800",
      "dispatches.jsonl": "cac2578d1b99c33f2f1486d38aa424f5b42caf5c77f9b277815f10be355f107b",
      "manifest.json": "40dd57f867e99dcc2c3688a35ec7ee9aff6c5e2d99576d0a9f2977d516db14d7",
      "model_metadata.json": "2ed475042b0f913ca45febe42aa811875069888f724f4389c6822209de357178",
      "probe_dispatch.json": "26eeabf5c7dcaa7886c030110f721996a2795e60ed57f7f33e286f0d71f0b84d",
      "results.jsonl": "cb92d8c1c505cf80b239f8c1f0d499938e522556b98301fa39a1b327a6669c42",
      "software.json": "a5ef9235c40b0b5c76199172365ed44d9bb938547fbe6e1822cd1c51b5e81faf"
    }
  },
  "scoring_authority_sha256": [
    "33a04e1b7ac8b3a3020423a37781e1ca5a3939081c6d350fd6a532a00bad77c2",
    "3b41037fba1cb02f42d0044a0dd758f1cc588de86fe0d2a5f90b142a41f353bb",
    "471b808d20d9be2608ef60fddbf8526bf9015af3d6a4d976b7a2b20843ad46e1",
    "4ca81b991c4002e70aead0d56b9f53c21f1e70d9b1fef1a4a1f2e32f32d5b56c",
    "4cd7c58fe4be67dadb3644fc54d46d8e4a40aaa8ed73c2005b6dbe52a4341a36",
    "8376ddce67cd9bb5f62d685d0309b4de5706f091084fbe72aed7dc91d1bf4dec",
    "aac4b2c2ece0f6cf37b81794031d93a8c3f53d032da1eb127d14f62be34781a6",
    "b5dba10a8c9fbcbd0fc79f62e11a9a08594fdc187cf3ead7b5bb872a43bf1b5f",
    "bb7ff1f81ac1b35bb93c25a5ac7dce48a0d86ff48cda711304b9932864c1bc3a",
    "c99db093ae3517dcacc03319f00a461acbfd19e0eb22cb0c5a5dd636f1ee9271",
    "ca5aa924ee10efe60bca09e02bb7fbb68740d12b243ca7bd619ceb5fc952c7bb",
    "dbac830fe3cad5f469f950fe2c3b1c2e575cb6d00c2bdaec4b82c46f41b0b8ab",
    "f6f685ea96ebd626804931278e03256bd502f0dc811a8694d9417a4ad5b64c01",
    "fbe6f2b49ad7c99416fb45dc05b8b9db37a3850017d7bfef81644d09afe9bcd4"
  ],
  "analysis_source_sha256": {
    "analyze_decision_additions_2026_10_21.py": "b09c3a6ae746acfd6d1dd7faa60a495ec146e7835f7e12d1ef45c38ad8a883b4",
    "analyze_direct_decisions_2026_10_20.py": "2517c3e6f8d32d2a9e0d403d8fc7a02f883883c42a3e008f7453addd43384916",
    "analyze_frontier_reference.py": "157c5695d1bc745c1eff6926cea7d138b18b1e4cbe72e20becb03b681628ddad",
    "test_analyze_decision_additions_2026_10_21.py": "5ffe3aa036d8dc4f0d8494c00979191656b80790658bfc9c167f324334629f7c"
  },
  "controls": {
    "mammoth1": {
      "n": 20,
      "scored": {
        "n": 20,
        "correct": 17,
        "answered": 20,
        "failed": 0,
        "wrong_valid": 3,
        "accuracy": 0.85
      },
      "direct_scored_same_subset": {
        "n": 20,
        "correct": 17,
        "answered": 20,
        "failed": 0,
        "wrong_valid": 3,
        "accuracy": 0.85
      },
      "same_choice": 20,
      "latency": {
        "n": 20,
        "median_ms": 232.29581299528945,
        "p95_ms": 483.9120098011339
      },
      "scope": "Separate20-case route control only; never pooled into full-bank direct results"
    },
    "elephant1": {
      "n": 20,
      "scored": {
        "n": 20,
        "correct": 13,
        "answered": 20,
        "failed": 0,
        "wrong_valid": 7,
        "accuracy": 0.65
      },
      "direct_scored_same_subset": {
        "n": 20,
        "correct": 13,
        "answered": 20,
        "failed": 0,
        "wrong_valid": 7,
        "accuracy": 0.65
      },
      "same_choice": 20,
      "latency": {
        "n": 20,
        "median_ms": 173.11102050734917,
        "p95_ms": 221.31295625076746
      },
      "scope": "Separate20-case route control only; never pooled into full-bank direct results"
    }
  },
  "verification": "Reconciled dispatches, attempts, results, software seals and every payload hash; private revision-bound scoring, raw first-attempt latency quantiles and charge sums verified. No route pooling.",
  "privacy": "Aggregate evidence and hashes only; no correct answer keys, per-item correctness, credentials or account records"
}
