Mathematical Reasoning on AIME 2024 (Accuracy@8)
87Accuracy (Avg@8)CERL full
Evaluation results
| Method | Links | |
|---|---|---|
| CERL fullTraining=SFT+RL2026.05 | 87 | |
| Fixed-ratio erasure + free answerTraining=None2026.05 | 83.7 | |
| HaloTraining=SFT+RL2026.05 | 83 | |
| Base Qwen3-32BTraining=N/A2026.05 | 81.5 | |
| Length-penalty RLTraining=RL2026.05 | 78.8 | |
| TokenSkip†Training=SFT2026.05 | 76.3 |