LLLin000
05e75e1b29
fix: PR1 deterministic OCR residual fixes
...
- Clamp backfill words to original bbox (Issue 3)
- Allow validation-first bare tables to fall through (Issue 1A)
- Materialize split table caption continuations (Issue 1B)
- Add short-form OCR health profile (Issue 4)
2026-07-01 21:49:45 +08:00
Research Assistant
4cd0cf3d54
fix: heading detection, backmatter, table consumed, footnote reorder, figure crop + audit fixtures
2026-06-22 01:28:55 +08:00
Research Assistant
70a7978de8
feat: complete OCR-v2 readiness gates + blind audit validation
...
- Gate 1: completeness signals for page text coverage
- Gate 2: DW figure ownership generalization with strict regression
- Gate 3: reference boundary enforcement from structure
- Gate 4: layout coverage taxonomy in audit/coverage_ledger.json
- diff_audit.py: canonical role alias mapping + auto-normalize truth_role
- ocr_structural_gate.py: table caption text-evidence fallback
- ocr-canonical-roles.md: canonical role list for audit workflow
- ocr-truth-audit.md: pre-flight checklist enforces canonical roles
- Blind audit: 5 unseen papers across domains, all PASS with coverage
2026-06-19 00:35:21 +08:00
Research Assistant
ccd08abb58
test: add synthetic PDF unit tests for pdf text fallback
2026-06-17 15:41:10 +08:00
