← Phase 1 / 08 · Small-Sample Version Regression: 25 → 12

Phase 1 · 08 v0.2 · 01 2026-08-12 Last Document of Today's Third Round

Small-Sample Version Regression: 25 → 12

The old fixture (2026-05-22) has a total of 25 curves, of which 10 are CLZ20 and 15 are Zha16; the current fixture (2026-06-03) has a total of 12 curves, of which 7 are CLZ20 and 5 are Zha16. Exact diff: kept 12, removed 13, added 0—the labels of the 13 removed curves are 106a1, 110b1, 110c1, 142d1, 142e1, 14a1, 26a1, 26b1, 34a1, 35a1, 38a1, 38b1, 66c1. The core discipline of the document is in the final paragraph: this diff can only state that a certain curve "changed from PASS in the old version to non-PASS in the current version"; before a filter-by-filter replay, they must all be marked as VERSION_REGRESSION_REMOVED, reason = OPEN, and mathematical reasons cannot be filled in from commit messages, branch distributions, or intuition—this is exactly a concrete instance of the distinction between Corpus A and Corpus B in document 07: membership change (this diff) does not equal theorem-level rejection reasons (which requires the official discrepancy_report.txt to list them predicate by predicate).

Removal list: 106a1, 110b1, 110c1, 142d1, 142e1, 14a1, 26a1, 26b1, 34a1, 35a1, 38a1, 38b1, 66c1——reason = OPEN — The phased status self-reported by the documents in the package, reproduced as is.

Connections · Connections

Relationship with other documents, try to use the words from its own document, not my interpretation.

Phase 1 Progress9 / 24 (Today's third round, and also today's final document)
"Before a filter-by-filter replay, always label as: VERSION_REGRESSION_REMOVED, reason = OPEN. You cannot fill in the mathematical reasons from commit messages, branch distributions, or intuition." — Excerpt from this document's "Do not over-interpret".

Loading...