Omission Detection on AbsenceBench (val)
98.63Poetry Micro-F1Claude 3.5 Sonnet
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Claude 3.5 SonnetPipeline=PDF→code (direct), Alias=paper_only2026.05 | 98.63 | 100 | 44.65 | 81.09 | 96.06 | 63.3 | |
| Claude 3.5 SonnetPipeline=PDF→CT→code, Alias=ct_only2026.05 | 98.62 | 100 | 23.36 | 73.99 | 91.58 | 61.41 | |
| Claude 3.7 Sonnet (thinking)Source=Reported2026.05 | 72.7 | 96 | 40 | 69.6 | — | — |