Math reasoning on AMC (out-of-distribution)
32.08AccuracyCalibRL
Evaluation Results
| Method | Links | |
|---|---|---|
| CalibRLTraining Data=9k-sample subset of MATH dataset2026.02 | 32.08 | |
| Qwen2.5-VL-7BTraining Data=9k-sample subset of MATH dataset2026.02 | 31.97 | |
| GRPOTraining Data=9k-sample subset of MATH dataset2026.02 | 31.59 |