Mathematical Problem Solving on Minerva (Pass@1)
33.32Pass@1VHG (Soft)
Evaluation Results
| Method | Links | |
|---|---|---|
| VHG (Soft)2026.05 | 33.32 | |
| Vanilla-GRPOtraining_variant=w. SFT data2026.05 | 32.63 | |
| Vanilla-GRPO2026.05 | 31.92 | |
| R-ZeroIteration=22026.05 | 28.52 | |
| R-ZeroIteration=32026.05 | 28.12 | |
| Qwen3-4B-Base2026.05 | 27.78 | |
| R-ZeroIteration=12026.05 | 23.74 |