Mathematical Reasoning on GSM8K In-Distribution (test)
91.8AccuracyCausalRM
Evaluation Results
| Method | Links | |
|---|---|---|
| CausalRMMethod Category=Proposed, Evaluation Type=Downstream RLHF2026.01 | 91.8 | |
| GoalRMMethod Category=Baseline, Evaluation Type=Downstream RLHF2026.01 | 89.4 | |
| Standard RMMethod Category=Baseline, Evaluation Type=Downstream RLHF2026.01 | 85.5 | |
| SFTMethod Category=Reference, Evaluation Type=Downstream RLHF2026.01 | 80.4 | |
| InfoRMMethod Category=Baseline, Evaluation Type=Downstream RLHF2026.01 | 71 |