Tool-Integrated Reasoning on AIME 24 (test accuracy)
43.85Test AccuracySeq-ALP
Evaluation Results
| Method | Links | |
|---|---|---|
| Seq-ALPtemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 43.85 | |
| Token-MIStemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 39.48 | |
| Seq-MIStemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 39.48 | |
| Token-ALPtemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 38.65 | |
| Seq-Bypasstemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 37.19 | |
| GRPOtemperature=1.0, sampling=average@32, multi-turn=true2026.03 | 35.42 |