Mathematical Reasoning on AIME 24 (Pass@1, Avg tokens)
12Pass@1 RatePROGRS-8
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PROGRS-8rollout_budget=82026.02 | 12 | 1,082.42 | |
| PROGRS-8 (No Centering)rollout_budget=8, centering=none2026.02 | 12 | 1,147.96 | |
| PROGRS-4rollout_budget=42026.02 | 10.67 | 1,201.29 | |
| PROGRS-8 (αcoh = 0)rollout_budget=8, coherence_penalty=disabled2026.02 | 9.33 | 1,018.21 | |
| DAPO-16rollout_budget=162026.02 | 8.67 | 1,202.22 | |
| DAPO-8rollout_budget=82026.02 | 8.67 | 1,245.09 |