audit(holoindex): rerun TQ2/TQ3 on frozen corpus baseline#432
Open
Foundup wants to merge 2 commits into
Open
Conversation
Locked audit window results on corpus freeze (23,836 docs): TQ2 (pure int8 vs fp32): - top-1: 92.0% (PASS ≥90%) - top-5: 64.0% (FAIL ≥95%) - sentinels: 29/30 (FAIL) - Decision: HOLD_INT8 TQ3 (routed int8/fp32): - top-1: 94.7% (PASS ≥90%) - top-5: 75.3% (FAIL ≥95%) - sentinels: 29/30 (FAIL) - Decision: HOLD_ROUTING Production default remains HOLO_USE_TURBOQUANT=0. Re-frozen manifest includes wsp_287 (FOUNDUPOPS doc from PR #425). Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
…frozen TQ audits CFZ3: Corpus Hygiene and Sentinel Hardening Phase 1 Corpus hygiene: - Add exclusion rules to index_wsp_entries() for hidden directories, _backup paths, and /archive/ paths - Removes 129 polluting documents from navigation_wsp (3451 -> 3322) - Specifically excludes .consciousness_migration_backup/ content Sentinel hardening: - Fix ambiguous sentinel query "WSP 97 truth distinction protocol" - Replace with canonical "WSP 97 System Execution Prompting Protocol" - TQ3 sentinels now pass 30/30 (was failing) Test coverage: - Add test_cfz3_corpus_hygiene.py with 10 exclusion tests Audit results (frozen corpus): - TQ2: HOLD_INT8 (88.7% top-1, 63.3% top-5, 1 sentinel fail on vocab) - TQ3: HOLD_ROUTING (95.3% top-1 PASS, 74.7% top-5 FAIL, sentinels PASS) No production policy change: HOLO_USE_TURBOQUANT=0 remains default. WSP: WSP 97 (truthful state reporting), WSP 50 (pre-action verification) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Locked audit window TQ2/TQ3 re-run on frozen corpus baseline (23,836 docs).
Gate Results
Decisions
HOLD_INT8HOLD_ROUTINGHOLO_USE_TURBOQUANT=0(unchanged)Corpus Stability
wsp_287(FOUNDUPOPS doc from PR docs(foundupops): manifest discovery and FAM registry architecture #425)Test plan
🤖 Generated with Claude Code