Code Generation on CRUX
57.2Score (%)INTUITOR
Evaluation Results
| Method | Links | |
|---|---|---|
| INTUITORBackbone=Qwen2.5-7B, Training Set=MATH2026.05 | 57.2 | |
| VIGORBackbone=Qwen2.5-7B, Training Set=MATH2026.05 | 56.38 | |
| GT-RewardBackbone=Qwen2.5-7B, Training Set=MATH2026.05 | 55.5 | |
| PAPOBase Model=Qwen2.5-14B-Base2026.05 | 48.25 | |
| 80/20Base Model=Qwen2.5-14B-Base2026.05 | 47.12 | |
| DAPOBase Model=Qwen2.5-14B-Base2026.05 | 46.38 | |
| Ent-RegBase Model=Qwen2.5-14B-Base2026.05 | 46.25 | |
| VIGORBackbone=Qwen2.5-3B, Training Set=MATH2026.05 | 40 | |
| INTUITORBackbone=Qwen2.5-3B, Training Set=MATH2026.05 | 38.7 | |
| GT-RewardBackbone=Qwen2.5-3B, Training Set=MATH2026.05 | 33 | |
| PAPOBase Model=Qwen2.5-7B-Base2026.05 | 30.88 | |
| DAPOBase Model=Qwen2.5-7B-Base2026.05 | 30.63 | |
| Ent-RegBase Model=Qwen2.5-7B-Base2026.05 | 29.75 | |
| 80/20Base Model=Qwen2.5-7B-Base2026.05 | 28.62 | |
| BaseBase Model=Qwen2.5-14B-Base2026.05 | 28.38 | |
| Before RL (Base)Backbone=Qwen2.5-3B, Training Set=MATH2026.05 | 24.38 | |
| Before RL (Base)Backbone=Qwen2.5-7B, Training Set=MATH2026.05 | 17.38 | |
| BaseBase Model=Qwen2.5-7B-Base2026.05 | 16.25 |