Judge Accuracy on RMB-SafeRLHF (Target)
69AccuracySeed Rubric
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Seed RubricJudge Model=Qwen3-14B2026.02 | 69 | 0.127 | |
| Biased Rubric SearchJudge Model=Qwen3-14B2026.02 | 67.4 | 0.182 | |
| Seed RubricJudge Model=Qwen3-14B2026.02 | 67.3 | 0.022 | |
| Biased Rubric SearchJudge Model=Qwen3-14B2026.02 | 64.1 | 0.062 |