What's Changed
- Perturbation benchmark v2 (supersedes #41) by @dangng2004 in #62
- Conference study benchmark: ICLR accepted vs rejected by @dangng2004 in #58
- Add competitor benchmark support (Coarse, cost estimator) by @dangng2004 in #66
- Scale up perturbation benchmark: context modes + substantive verifier by @dangng2004 in #67
- Thread span offset through CandidateSpan and Perturbation by @dangng2004 in #64
- Improve review prompts and add --skip-nonsubstantial flag by @dangng2004 in #63
- Land PR #67 (perturbation-scaleup) into main by @dangng2004 in #70
- Generalized perturbation verifiers by @dangng2004 in #73
- Collapse runners into run_benchmark.py with system abstraction by @dangng2004 in #74
- Review analysis by @anitahao in #78
- conference_study: scale-up matrix, real OpenRouter costs, coarse page-cap by @dangng2004 in #79
- conference_study: add analyses scripts; gitignore study artifacts by @dangng2004 in #80
- docs: benchmarks README + expand top-level Benchmarks section by @dangng2004 in #81
- gitignore: ignore local conference_study configs except canonical three by @dangng2004 in #89
- gitignore: blanket-ignore benchmarks/perturbation/configs/ by @dangng2004 in #92
- Revert "Improve review prompts and add --skip-nonsubstantial flag" by @dangng2004 in #68
- Add cluster-bootstrap 95% CIs for the paper's AUC tables by @dangng2004 in #98
- Fix GLM/Kimi returning 0 comments by @r-uben in #55
- Pre-release cleanup: drop debug print, untrack venv/egg-info artifacts by @dangng2004 in #107
- Release 0.3.0 by @dangng2004 in #108
New Contributors
Full Changelog: v0.2.8...v0.3.0
Contributors
dangng2004, r-uben, and anitahao