Mathematical Reasoning on GSM8K, MATH500, OlympiadBench, AIME (test)
71.4Pass@1DRPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DRPOModel Size=7B2025.10 | 71.4 | 1,502 | 0.249 | |
| RLOO-LPModel Size=7B2025.10 | 69.2 | 2,649 | -0.033 | |
| DRPOModel Size=8B2025.10 | 69 | 1,758 | 0.297 | |
| ALPModel Size=7B2025.10 | 67.9 | 2,170 | -0.134 | |
| ALPModel Size=8B2025.10 | 63.7 | 2,358 | -0.01 | |
| DRPOModel Size=1.5B2025.10 | 62.4 | 1,527 | 0.178 | |
| RLOO-LPModel Size=8B2025.10 | 61.9 | 2,249 | 0.251 | |
| ALPModel Size=1.5B2025.10 | 60.6 | 3,494 | -0.387 | |
| HAPOModel Size=7B2025.10 | 59.6 | 1,717 | -1.08 | |
| HAPOModel Size=8B2025.10 | 59.5 | 1,981 | -0.366 | |
| RLOO-LPModel Size=1.5B2025.10 | 56.7 | 2,531 | -0.129 | |
| HAPOModel Size=1.5B2025.10 | 53.4 | 1,791 | -0.519 |