Mathematical Reasoning on GSM-Hard Out-of-Distribution (test)
71Final Answer AccuracyCausalRM
Evaluation Results
| Method | Links | |
|---|---|---|
| CausalRMEvaluation Type=Downstream RLHF2026.01 | 71 | |
| GoalRMEvaluation Type=Downstream RLHF2026.01 | 70.1 | |
| SFTEvaluation Type=Downstream RLHF2026.01 | 54.5 | |
| Standard RMEvaluation Type=Downstream RLHF2026.01 | 50.3 | |
| InfoRMEvaluation Type=Downstream RLHF2026.01 | 37.5 |