Tool-Integrated Reasoning on Minerva Math
43.18Test AccuracyToken-ALP
Evaluation Results
| Method | Links | |
|---|---|---|
| Token-ALPtemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 43.18 | |
| Seq-ALPtemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 43.1 | |
| GRPOtemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 42.13 | |
| Seq-MIStemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 42.05 | |
| Token-MIStemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 41.45 | |
| Seq-Bypasstemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 39.21 |