test(rectification): confirmation gate reads the upstream_sync4 re-frozen rerun (ERR-110)

Original / new / reason recorded in the test. Shadow trials identical; gate
behaviour and thresholds unchanged.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
This commit is contained in:
Jesse_Chen
2026-10-02 14:50:35 +08:00
co-authored by Claude Opus 5.5
parent 8ae4cef7a6
commit 26d4c204b0
@@ -90,8 +90,11 @@ const diagnosticReport = JSON.parse(readFileSync(
// 原值(3):docs/research/sealed_holdout_rerun_quality_wording_2026_09_29.json
// 新值(3):docs/research/sealed_holdout_rerun_functional_v2_2026_10_02.json
// 原因(3):BUG-1181 功能吉凶 v2 + scoring-10,按 ERR-110 另存新冻结重跑;影子 20 例逐条不变,确认门行为与阈值不变。
// 原值(4):docs/research/sealed_holdout_rerun_functional_v2_2026_10_02.json
// 新值(4):docs/research/sealed_holdout_rerun_upstream_sync4_2026_10_02.json
// 原因(4):BUG-1188/1189 改了冻结身份里的 scripts/shadbala.py,按 ERR-110 另存新冻结重跑;影子 20 例逐条不变,算法版本不变,确认门行为与阈值不变。
const currentTreeRerun = JSON.parse(readFileSync(
new URL("../../docs/research/sealed_holdout_rerun_functional_v2_2026_10_02.json", import.meta.url),
new URL("../../docs/research/sealed_holdout_rerun_upstream_sync4_2026_10_02.json", import.meta.url),
"utf8",
)) as {
frozen_record: { implementation_sha256: string };
@@ -232,9 +235,10 @@ test("sealed holdout aggregates match the v3 report and still block confirmation
);
// 原值:docs/research/sealed_holdout_rerun_midnight_anchor_2026_09_21.json / 新值:…quality_wording_2026_09_29.json / 原因:BUG-1088 重新冻结(同上)。
// 原值(2):…quality_wording_2026_09_29.json / 新值(2):…functional_v2_2026_10_02.json / 原因(2):BUG-1181 重新冻结(同上)。
// 原值(3):…functional_v2_2026_10_02.json / 新值(3):…upstream_sync4_2026_10_02.json / 原因(3):BUG-1188/1189 重新冻结(同上)。
assert.equal(
productHoldout.current_tree_scorer.source_report,
"docs/research/sealed_holdout_rerun_functional_v2_2026_10_02.json",
"docs/research/sealed_holdout_rerun_upstream_sync4_2026_10_02.json",
);
assert.equal(currentTreeRerun.official_valid_independent_blind, false);
assert.equal(productHoldout.current_tree_scorer.matches_metrics_scorer, false);