Mathematical Reasoning on AIME 2025 (average@32)
60.42Average Score (32 samples)GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GRPOParams=4B2026.05 | 60.42 | |
| SODParams=1.7B2026.05 | 41.72 | |
| OPDParams=1.7B2026.05 | 37.04 | |
| SODParams=0.6B2026.05 | 26.13 | |
| OPSDgtParams=1.7B2026.05 | 24.69 | |
| OPDParams=0.6B2026.05 | 22.95 | |
| SFTParams=1.7B2026.05 | 22.4 | |
| GRPOParams=1.7B2026.05 | 21.67 | |
| OPSDhintParams=1.7B2026.05 | 21.43 | |
| OPSDgtParams=0.6B2026.05 | 17.04 | |
| OPSDhintParams=0.6B2026.05 | 14.12 | |
| VanillaParams=0.6B2026.05 | 12.81 | |
| VanillaParams=1.7B2026.05 | 8.96 | |
| SFTParams=0.6B2026.05 | 5.42 | |
| GRPOParams=0.6B2026.05 | 4.9 |