Open Question Answering on AIME 2025 (test)
70AccuracyGRPO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GRPO2026.02 | 70 | 9,871 | |
| ESTARoptimization=RL2026.02 | 70 | 3,788 | |
| O1-Pruner2026.02 | 66.67 | 7,841 | |
| FlashThink2026.02 | 66.67 | 6,504 | |
| AdaptThink2026.02 | 66.67 | 4,513 | |
| Length-Penaltyexplicit_length_penalty=true2026.02 | 66.67 | 7,324 | |
| ESTAR-LITEearly_stopping=classifier-based2026.02 | 66.67 | 3,045 | |
| ESTAR-FTmode=fine-tuned2026.02 | 66.67 | 3,413 | |
| No-Thinkingreasoning=disabled2026.02 | 26.67 | 1,513 |