{
  "schema": "anvil-serving.independent-outcome-review/v1",
  "reviewer": "independent Sol",
  "decision": "The frozen local gate failed correctly; no conditional model probe.",
  "confirmed": [
    "Eight forced cases passed.",
    "Stock and patch masks were identical on 271 of 272 ordinary rows.",
    "All 136 holdout rows were identical between versions.",
    "The only changed row retained one extra token, 30317, with binary64 reference mass 2.0449949975313364e-6.",
    "All reported input/output bit-integrity and prefix/subset checks passed.",
    "The change is within the predeclared count envelope derived from the pinned upstream tests; this does not establish that the upstream test suite passed."
  ],
  "not_proven": [
    "The end-to-end reasoning limit is repaired.",
    "A unique internal pivot or duplicate-handling cause; no internal pivot was captured.",
    "Any performance or quality impact."
  ],
  "next_experiment": "In a new frozen plan, separate an upstream-compatible policy from exact minimality or capture pivot, minimum-larger value, and duplicate budget for the single frozen row. Do not retroactively pass or rerun this attempt.",
  "promoted": false
}