Math Reasoning on GSM8K (Pass@K)
83.2Pass@1 AccuracyMIG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MIGEvaluation Mode=In-Domain, Training Strategy=Task-Specific Training2026.02 | 83.2 | 95.6 | 2.2 | |
| GRPOEvaluation Mode=In-Domain, Training Strategy=Task-Specific Training2026.02 | 81 | 95.6 | — | |
| Base ModelEvaluation Mode=In-Domain, Training Strategy=Task-Specific Training2026.02 | 77.8 | 93.2 | — |