Math Robustness on SVAMP
88.67Pass@1MIG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MIGEvaluation Mode=Out-of-Distribution (OOD), Training Strategy=Transferred from MATH (300 steps)2026.02 | 88.67 | 96.67 | 3.67 | |
| Base ModelEvaluation Mode=Out-of-Distribution (OOD), Training Strategy=Transferred from MATH (300 steps)2026.02 | 85 | 96.67 | — | |
| GRPOEvaluation Mode=Out-of-Distribution (OOD), Training Strategy=Transferred from MATH (300 steps)2026.02 | 85 | 97.67 | — |