List your company

Judge Accuracy on RMB-SafeRLHF (Bench)

85.6AccuracyBiased Rubric Search
68.85673.20377.5581.897Feb 14, 2026

Evaluation results

MethodLinks
Biased Rubric SearchJudge Model=Qwen3-14B2026.0285.6
Seed RubricJudge Model=Qwen3-14B2026.0281.7
Biased Rubric SearchJudge Model=Qwen3-14B2026.0270.3
Seed RubricJudge Model=Qwen3-14B2026.0269.5