Mathematical Reasoning on 5 O.O.D. Datasets Averaged
52.5AccuracyRLVR
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| RLVRTraining Set=Big-Math2025.07 | 52.5 | 0.52 | 0.49 | 0.49 | |
| RLVR + BCE ClassifierTraining Set=Big-Math2025.07 | 52.5 | 0.55 | 0.34 | 0.33 | |
| RLVR + Brier ClassifierTraining Set=Big-Math2025.07 | 52.5 | 0.57 | 0.28 | 0.27 | |
| RLVR + ProbeTraining Set=Big-Math2025.07 | 52.5 | 0.53 | 0.33 | 0.3 | |
| Answer ProbabilityTraining Set=Big-Math2025.07 | 52.5 | 0.52 | 0.44 | 0.43 | |
| RLCRTraining Set=Big-Math2025.07 | 50.9 | 0.6 | 0.28 | 0.25 | |
| BaseTraining Set=Big-Math2025.07 | 47.8 | 0.53 | 0.46 | 0.45 | |
| SFT+RLCRTraining Set=Big-Math2025.07 | 43.8 | 0.66 | 0.24 | 0.18 |