Multi-step narrative inference on MuSR
68.76Exact MatchQUBRIC
Evaluation results
| Method | Links | |
|---|---|---|
| QUBRIC2026.06 | 68.76 | |
| Math-RLVRLength Penalty=0.52026.06 | 65.46 | |
| AutoIF-RLVRTraining Domain=IF-domain2026.06 | 65.36 | |
| Rubric-RLQuery=Original Query2026.06 | 64.73 | |
| SFT Baseline2026.06 | 64.51 | |
| Math-RLVRLength Penalty=none2026.06 | 52.1 |