Conformal Risk Control on Aggregate (CoNLL, FNERD, WNUT, TQA, NQ, STEM, Humanities, JSON) (test)
0.009RiskCRC
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CRCalpha=0.10, formal_guarantee=true2026.06 | 0.009 | 0.075 | 0 | |
| SC Thr. (no guar.)alpha=0.10, formal_guarantee=false2026.06 | 0.019 | 0.846 | 7.5 | |
| Entropy Thr.alpha=0.102026.06 | 0.021 | 0.868 | 12.5 | |
| TM Thr. (no guar.)alpha=0.10, formal_guarantee=false2026.06 | 0.022 | 0.869 | 7.5 | |
| Always Answeralpha=0.102026.06 | 0.434 | 0 | 90 |