Mathematical Problem Solving on AIME 2025 (Pass@12, Avg@12, Maj@12)
83.33Pass@12OPSD
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OPSDModel=Qwen3-4B, Max Generation Length=38,912 tokens, Number of Evaluation Seeds=32026.05 | 83.33 | 66.67 | 73.33 | |
| PW-OPSD ModerateModel=Qwen3-4B, Max Generation Length=38,912 tokens, Number of Evaluation Seeds=3, Schedule Type=Moderate2026.05 | 83.33 | 67.78 | 74.44 | |
| PW-OPSD AggressiveModel=Qwen3-4B, Max Generation Length=38,912 tokens, Number of Evaluation Seeds=3, Schedule Type=Aggressive2026.05 | 83.33 | 67.59 | 75.56 | |
| EOPDModel=Qwen3-4B, Max Generation Length=38,912 tokens, Number of Evaluation Seeds=32026.05 | 82.22 | 65.65 | 76.67 | |
| REOPOLDModel=Qwen3-4B, Max Generation Length=38,912 tokens, Number of Evaluation Seeds=32026.05 | 81.11 | 62.13 | 74.44 |