Mathematical Problem Solving on AIME 2026 (Pass@1 %)
12.92Pass@1VHG (Soft)
Evaluation Results
| Method | Links | |
|---|---|---|
| VHG (Soft)2026.05 | 12.92 | |
| Vanilla-GRPOtraining_variant=w. SFT data2026.05 | 9.58 | |
| Vanilla-GRPO2026.05 | 8.12 | |
| Qwen3-4B-Base2026.05 | 7.71 | |
| R-ZeroIteration=22026.05 | 7.5 | |
| R-ZeroIteration=32026.05 | 6.67 | |
| R-ZeroIteration=12026.05 | 6.25 |