Adding Mistake on MuSR
0.731AOCFaithfulness Only
Evaluation Results
| Method | Links | |
|---|---|---|
| Faithfulness OnlyBackbone=Qwen3-14B2026.02 | 0.731 | |
| MAT-SteerBackbone=Qwen3-14B2026.02 | 0.722 | |
| REMULBackbone=Qwen3-14B2026.02 | 0.722 | |
| Balanced RewardsBackbone=Qwen3-14B2026.02 | 0.712 | |
| OriginalBackbone=Qwen3-14B2026.02 | 0.708 | |
| Correctness OnlyBackbone=Qwen3-14B2026.02 | 0.694 | |
| Hint OptimizedBackbone=Qwen3-14B2026.02 | 0.688 |