List your company

Judge Accuracy on Ultra-Problem (Bench)

73AccuracyBiased Rubric Search
72.79272.84672.972.954Feb 14, 2026

Evaluation results

MethodLinks
Biased Rubric SearchJudge Model=Qwen3-14B2026.0273
Seed RubricJudge Model=Qwen3-14B2026.0272.8