Mathematical Problem Solving on AIME 2024 (Pass@1 %)
13.96Pass@1 AccuracyVanilla-GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| Vanilla-GRPO2026.05 | 13.96 | |
| VHG (Soft)2026.05 | 13.12 | |
| Vanilla-GRPOtraining_variant=w. SFT data2026.05 | 11.04 | |
| R-ZeroIteration=32026.05 | 10.83 | |
| R-ZeroIteration=12026.05 | 10.62 | |
| R-ZeroIteration=22026.05 | 10 | |
| Qwen3-4B-Base2026.05 | 7.29 |