Mathematical Reasoning on MATH (test) (Pass@1, Pass@8)
74.2Pass@1Sequence-level adapter
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Sequence-level adapterSetting=mix CoT, Budget=w/ budget2026.03 | 74.2 | 79.8 | |
| Sequence-level adapterSetting=mix CoT, Budget=w/o budget2026.03 | 73.6 | 78.8 | |
| Sequence-level adapterSetting=w/o CoT, Budget=w/ budget2026.03 | 72.9 | 78.46 | |
| TURNSetting=TURN2026.03 | 72.72 | 76.96 | |
| Sequence-level adapterSetting=w/o CoT, Budget=w/o budget2026.03 | 72.6 | 78.3 | |
| Static sampling (Best)Setting=mix CoT, Sampling Strategy=Best2026.03 | 72.1 | 77.1 | |
| Static sampling (Mixed)Setting=mix CoT, Sampling Strategy=Mixed2026.03 | 71.93 | 76.57 | |
| Static sampling (Best)Setting=w/o CoT, Sampling Strategy=Best2026.03 | 71.7 | 76.7 | |
| Static sampling (Mixed)Setting=w/o CoT, Sampling Strategy=Mixed2026.03 | 71.2 | 76.23 |