Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Preference evaluation on Ultra-Real benchmark

43.1Win Rate

πbias (rubric-based preference attack)

39.56440.48241.442.318Feb 14, 2026
Updated 4d ago

Evaluation Results

MethodLinks
2026.02
43.1
2026.02
39.7