List your company

Preference Alignment Evaluation on PKU-SafeRLHF

83Win Rate (Llama Judge)Pref-CTRL
77.879.1580.581.85Apr 26, 2026

Evaluation results

MethodLinks
Pref-CTRLTrained Dataset=SHP, Zero-shot transfer setting=true2026.048375
RE-ControlTrained Dataset=SHP, Zero-shot transfer setting=true2026.048173
Pref-CTRLTrained Dataset=HH-RLHF, Zero-shot transfer setting=true2026.048067
RE-ControlTrained Dataset=HH-RLHF, Zero-shot transfer setting=true2026.047865