Mathematical Reasoning on AIME 25 (Accuracy (Avg@8))
75.7Accuracy (Avg@8)CERL full
Evaluation Results
| Method | Links | |
|---|---|---|
| CERL fullTraining=SFT+RL2026.05 | 75.7 | |
| HaloTraining=SFT+RL2026.05 | 73.2 | |
| Fixed-ratio erasure + free answerTraining=None2026.05 | 72.2 | |
| Base Qwen3-14BTraining=N/A2026.05 | 70.5 | |
| Length-penalty RLTraining=RL2026.05 | 69.4 | |
| TokenSkip†Training=SFT2026.05 | 65.5 |