Mathematical Reasoning on HMMT F25 (avg@16, pass@16)
11.25Avg@16 ScoreRollout-level |Ai| replay
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Rollout-level |Ai| replayModel=Qwen3-4B2026.06 | 11.25 | 36.67 | |
| uniformModel=Qwen3-4B2026.06 | 7.81 | 22.15 | |
| GRPOModel=Qwen3-4B2026.06 | 7.29 | 16.67 | |
| σgModel=Qwen3-4B2026.06 | 7.08 | 30 | |
| Rollout-level |Ai| replayModel=Qwen3-1.7B2026.06 | 0.62 | 6.67 | |
| uniformModel=Qwen3-1.7B2026.06 | 0.52 | 6.29 | |
| GRPOModel=Qwen3-0.6B2026.06 | 0.42 | 3.33 | |
| Rollout-level |Ai| replayModel=Qwen3-0.6B2026.06 | 0.42 | 3.33 | |
| σgModel=Qwen3-0.6B2026.06 | 0.21 | 3.33 | |
| GRPOModel=Qwen3-1.7B2026.06 | 0.21 | 3.33 | |
| σgModel=Qwen3-1.7B2026.06 | 0.21 | 3.33 | |
| uniformModel=Qwen3-0.6B2026.06 | 0 | 0 |