Calibration on HotPotQA Hard
0.13Top-1 ECERLCR Single
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RLCR SingleRL Algorithm=RLCR, Output Mode=Single, k=32026.03 | 0.13 | — | — | 0.16 | — | |
| RLCR SingleAnswer Setting=Single, k=32026.03 | 0.13 | — | — | 0.16 | — | |
| RLCR MultiRL Algorithm=RLCR, Output Mode=Multi, k=32026.03 | 0.31 | 0.44 | 0.21 | 0.22 | 0.12 | |
| RLCR MultiAnswer Setting=Multi, k=32026.03 | 0.31 | 0.44 | 0.21 | 0.22 | 0.12 | |
| RLVR MultiRL Algorithm=RLVR, Output Mode=Multi, k=32026.03 | 0.38 | 0.47 | 0.22 | 0.3 | 0.16 | |
| RLVR MultiAnswer Setting=Multi, k=32026.03 | 0.38 | 0.47 | 0.22 | 0.3 | 0.16 | |
| RLVR SingleRL Algorithm=RLVR, Output Mode=Single, k=32026.03 | 0.48 | — | — | 0.38 | — | |
| RLVR SingleAnswer Setting=Single, k=32026.03 | 0.48 | — | — | 0.38 | — |