Multi-task Language Understanding on MMLU (Best-of-K Accuracy)
85.7Best-of-K AccuracyVeriBound
Evaluation Results
| Method | Links | |
|---|---|---|
| VeriBoundK=5, Evaluation Mode=Empirical2026.06 | 85.7 | |
| FOVERK=52026.06 | 85.1 | |
| ReasonFlux-PRMK=52026.06 | 84.3 | |
| R-PRMK=52026.06 | 83.9 | |
| VeriBoundK=5, Evaluation Mode=Bound2026.06 | 83.8 | |
| PRM800KK=52026.06 | 83.2 | |
| Math-ShepherdK=52026.06 | 81.3 | |
| ORMK=52026.06 | 78.5 |