{
  "date": "2026-09-09",
  "scope": "Same frozen 160-case document/retention board and document-search prompt; not the historical 240-case filename board",
  "incumbent": {
    "attention": "sdpa",
    "board_sha256": "d6056f70595906548257e3fa08cf391c1285b59ec2ef9506f00c06c7cb0e8def",
    "device": "mps",
    "dtype": "bfloat16",
    "families": {
      "all_any": {
        "n": 10,
        "outcome_exact": 0,
        "plan_exact": 0
      },
      "capability": {
        "n": 10,
        "outcome_exact": 2,
        "plan_exact": 2
      },
      "compound": {
        "n": 10,
        "outcome_exact": 0,
        "plan_exact": 0
      },
      "credential": {
        "n": 10,
        "outcome_exact": 0,
        "plan_exact": 0
      },
      "filename": {
        "n": 10,
        "outcome_exact": 10,
        "plan_exact": 0
      },
      "fresh_all": {
        "n": 40,
        "outcome_exact": 0,
        "plan_exact": 0
      },
      "fresh_any": {
        "n": 40,
        "outcome_exact": 0,
        "plan_exact": 0
      },
      "or_recall": {
        "n": 10,
        "outcome_exact": 0,
        "plan_exact": 0
      },
      "scope_format": {
        "n": 10,
        "outcome_exact": 0,
        "plan_exact": 0
      },
      "window_topk": {
        "n": 10,
        "outcome_exact": 0,
        "plan_exact": 0
      }
    },
    "host": "local Mac laptop",
    "human_evidence": false,
    "model_sha256": "8ccf00779bf93e16981c6f0272197a768165c6a21a6c8b5e4f81a9b1f9371ef0",
    "outcome_exact": 12,
    "plan_exact": 2,
    "plan_scored": 150,
    "rows": 160
  },
  "candidate": {
    "attention": "sdpa",
    "board_sha256": "d6056f70595906548257e3fa08cf391c1285b59ec2ef9506f00c06c7cb0e8def",
    "device": "mps",
    "dtype": "bfloat16",
    "families": {
      "all_any": {
        "n": 10,
        "outcome_exact": 10,
        "plan_exact": 10
      },
      "capability": {
        "n": 10,
        "outcome_exact": 9,
        "plan_exact": 9
      },
      "compound": {
        "n": 10,
        "outcome_exact": 10,
        "plan_exact": 10
      },
      "credential": {
        "n": 10,
        "outcome_exact": 10,
        "plan_exact": 10
      },
      "filename": {
        "n": 10,
        "outcome_exact": 10,
        "plan_exact": 0
      },
      "fresh_all": {
        "n": 40,
        "outcome_exact": 37,
        "plan_exact": 32
      },
      "fresh_any": {
        "n": 40,
        "outcome_exact": 34,
        "plan_exact": 34
      },
      "or_recall": {
        "n": 10,
        "outcome_exact": 6,
        "plan_exact": 6
      },
      "scope_format": {
        "n": 10,
        "outcome_exact": 10,
        "plan_exact": 10
      },
      "window_topk": {
        "n": 10,
        "outcome_exact": 10,
        "plan_exact": 10
      }
    },
    "host": "local Mac laptop",
    "human_evidence": false,
    "model_sha256": "984ff1aaed71652d9f53c91074e0787b3386bc3ae41900b4b830e6211805fc64",
    "outcome_exact": 146,
    "plan_exact": 131,
    "plan_scored": 150,
    "rows": 160
  },
  "change_counts": {
    "outcome_improvements": 134,
    "outcome_regressions": 0,
    "plan_improvements": 129,
    "plan_regressions": 0
  },
  "sandboxed_broker_replay": {
    "gold_passed": 160,
    "incumbent_outcomes": 12,
    "candidate_outcomes": 146,
    "outcome_regressions": 0,
    "binary_sha256": "f87a44fc86fbe4637a435b00f17c10be6c604a46496f5f5625c3014522ae1b47",
    "runtime": "Actual Darwin broker with unchanged sandbox and approval policy; saved generations using document-search prompt"
  },
  "candidate_predictions_reused_after_provenance_verification": true,
  "operator_release_readiness_criterion": "Candidate outcomes greater than incumbent on this comparison",
  "operator_release_readiness_criterion_met": true,
  "original_saved_a_gate_passed": false,
  "promoted": false,
  "installed_or_published": false,
  "synthetic_evidence_only": true,
  "legacy_240_case_board_rerun": false
}
