Mathematical Reasoning on AIME24 (Pass@1 Accuracy)
83Pass@1 AccuracyTeacher (GRPO)
Evaluation results
| Method | Links | |
|---|---|---|
| Teacher (GRPO)Model Family=Qwen Family2026.06 | 83 | |
| PADDModel Family=Qwen Family2026.06 | 83 | |
| GSPOModel Family=Qwen Family2026.06 | 80.4 | |
| RSPOModel Family=Qwen Family2026.06 | 80.3 | |
| Online KDModel Family=Qwen Family2026.06 | 78.4 | |
| Vanilla-GRPOModel Family=Qwen Family2026.06 | 76.8 | |
| BaseModel Family=Qwen Family2026.06 | 74.5 | |
| PADDModel Family=DeepSeek Family2026.06 | 57.6 | |
| RSPOModel Family=DeepSeek Family2026.06 | 55.7 | |
| Teacher (GRPO)Model Family=DeepSeek Family2026.06 | 55.3 | |
| GSPOModel Family=DeepSeek Family2026.06 | 54.8 | |
| Vanilla-GRPOModel Family=DeepSeek Family2026.06 | 50.3 | |
| Online KDModel Family=DeepSeek Family2026.06 | 50.1 | |
| Dense-GRPOModel Family=DeepSeek Family2026.06 | 46.2 | |
| Dense-GRPOModel Family=Qwen Family2026.06 | 40.1 | |
| BaseModel Family=DeepSeek Family2026.06 | 38.4 |