Reasoning on MoReBench
56.59LLM-as-a-judge ScoreQUBRIC
Evaluation Results
| Method | Links | |
|---|---|---|
| QUBRIC2026.06 | 56.59 | |
| Rubric-RLQuery=Original Query2026.06 | 53.1 | |
| Math-RLVRLength Penalty=none2026.06 | 53.01 | |
| AutoIF-RLVRTraining Domain=IF-domain2026.06 | 52.31 | |
| SFT Baseline2026.06 | 50.74 | |
| Math-RLVRLength Penalty=0.52026.06 | 48.74 |