Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Preference Alignment Evaluation on PKU-SafeRLHF
Loading...
83
Win Rate (Llama Judge)
Pref-CTRL
77.8
79.15
80.5
81.85
Apr 26, 2026
Win Rate (Llama Judge)
Win Rate (DeepSeek Judge)
Updated 2mo ago
Evaluation Results
Method
Method
Links
Win Rate (Llama Judge)
Win Rate (DeepSeek Judge)
Pref-CTRL
Trained Dataset=SHP, Z...
2026.04
83
75
RE-Control
Trained Dataset=SHP, Z...
2026.04
81
73
Pref-CTRL
Trained Dataset=HH-RLH...
2026.04
80
67
RE-Control
Trained Dataset=HH-RLH...
2026.04
78
65
Feedback
Search any
task
Search any
task