Skip to content

Release 0.2.2: corrected benchmarks with latest competitor versions#9

Merged
jdrhyne merged 2 commits intomainfrom
release/0.2.2
Apr 24, 2026
Merged

Release 0.2.2: corrected benchmarks with latest competitor versions#9
jdrhyne merged 2 commits intomainfrom
release/0.2.2

Conversation

@jdrhyne
Copy link
Copy Markdown
Contributor

@jdrhyne jdrhyne commented Apr 23, 2026

Summary

Re-ran benchmarks with all competitor libraries updated to their latest versions and ODL hybrid properly configured (requires running docling server). This fixes scoring anomalies flagged by Matej.

What changed

  • opendataloader-pdf 1.9.1 → 2.3.0
  • docling 2.71.0 → 2.91.0
  • pymupdf4llm 0.3.4 → 1.27.2 (major version, quality recovered)
  • markitdown 0.1.4 → 0.1.5 (table extraction restored: TEDS 0.00 → 0.27)
  • ODL hybrid now runs with docling server (TEDS 0.43 → 0.68, overall 0.83 → 0.87)
  • Exact library versions now shown in benchmark tables
  • Chart images refreshed
  • Version bump 0.2.1 → 0.2.2

Nutrient results unchanged

Test plan

  • Verify benchmark tables match April 23 evaluation.json files
  • Verify chart images render correctly on GitHub
  • After merge: npm publish --access public

jdrhyne added 2 commits April 23, 2026 11:29
- Re-run benchmarks with all competitors on latest versions
- Pin exact library versions in tables for transparency
- Fix ODL hybrid (was running without server, now properly configured)
- Key version bumps: pymupdf4llm 0.3.4→1.27.2, opendataloader 1.9.1→2.3.0,
  docling 2.71.0→2.91.0, markitdown 0.1.4→0.1.5
- Refresh chart images from April 23 run
- Bump version to 0.2.2
@jdrhyne jdrhyne merged commit b7cf844 into main Apr 24, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant