List your company

Judge Accuracy on Ultra-Problem (Target)

61.5AccuracySeed Rubric
57.44458.49759.5560.603Feb 14, 2026

Evaluation results

MethodLinks
Seed RubricJudge Model=Qwen3-14B2026.0261.511.3
Biased Rubric SearchJudge Model=Qwen3-14B2026.0257.615.4