List your company

Preference Alignment on K-Sort Eval

21.63K-Sort Eval ScoreArenaPO
19.96620.39820.8321.262May 7, 2026

Evaluation results

MethodLinks
ArenaPO2026.0521.63
Fixed-margin baselineMargin=2.02026.0521.16
Fixed-margin baselineMargin=4.02026.0520.95
Fixed-margin baselineMargin=0.52026.0520.72
Stable Diffusion 1.52026.0520.03