List your company

Helpfulness preference labeling accuracy on Ultra-Creative

73.7Accuracy (Bench)Few-Shot
70.16471.0827272.918Feb 14, 2026

Evaluation results

MethodLinks
Few-ShotJudge=Deepseek-V32026.0273.758.615.1
Few-ShotJudge=Qwen3-14B2026.0273.557.815.7
Biased Rubric SearchJudge=Deepseek-V32026.0273.554.718.8
RandomJudge=Qwen3-14B2026.0273.456.217.2
SeedJudge=Deepseek-V32026.0273.459.613.8
RandomJudge=Deepseek-V32026.0273.458.514.9
TextGradJudge=Deepseek-V32026.0273.460.512.9
SeedJudge=Qwen3-14B2026.0272.859.113.7
Few-ShotJudge=Gemma-3-27b-it2026.0272.759.313.4
Biased Rubric SearchJudge=Qwen3-14B2026.0272.154.517.6
TextGradJudge=Qwen3-14B2026.027257.314.7
RandomJudge=Gemma-3-27b-it2026.0271.55813.5
TextGradJudge=Gemma-3-27b-it2026.0271.558.213.3
Biased Rubric SearchJudge=Gemma-3-27b-it2026.02715813
SeedJudge=Gemma-3-27b-it2026.0270.360.110.2