- Fix references zone detection without heading (backward reference fallback)
- Fix reference_item silently dropped in _reorder_tail_run_fifo when no heading
- Fix figure proximity matching: use nearest legend regardless of column order
- Fix figure numbering in extract_and_write_objects: use actual figure_id
- Fix body→reference_item rescue: require text starts with numeric ref pattern
- Fix tail reading order rebuilt after block normalization
- A8E7SRVS: corrected figure-legend pairing on page 5
- CAQNW9Q2: references now render without heading
- K7R8PEKW: adjusted body retention threshold
- Add RegionPrepass data model for frontmatter/structured_insert/body classification
- Add PDF visual container detection (filled rects) for sidebar detection
- Add Box N and keyword-based sidebar anchor detection
- Add column-aware block expansion for mixed sidebar clusters
- Add _container_text from PDF text blocks for cross-column truncation
- Add degraded-mode heading level normalization with font size clustering
- Merge adjacent structured_insert blocks into single callout
- Fix metadata enrichment from Literature-hub note frontmatter for old papers
- Fix heading hierarchy using span font-size clusters when available
- Add real-paper regression harness (TSCKAVIS, CAQNW9Q2, A8E7SRVS, K7R8PEKW)
- Add body retention, heading, metadata, and callout acceptance tests
- 219 passed, 4 control guards passed
Block 35 in SAN9AYVR was a biography continuation fragment that escaped
non_body_insert detection because it shared the same width and font as
the body spine. Added a second-pass expansion in
_detect_non_body_insert_clusters that catches orphan continuation
fragments (lowercase start, same font as insert cluster, adjacent to
detected insert) and marks them as non_body_insert.
Regression test: test_non_body_insert_catches_continuation_fragment
- Expand role filter in _detect_non_body_insert_clusters from body_paragraph
only to include frontmatter_noise and unknown_structural
- Replace absolute page<=3 gate with relative bound based on body_end_page:
max_early_page = min(3, body_end_page/4 + 1)
- Added test_detect_non_body_insert_catches_frontmatter_noise_blocks
