Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Harmlessness Evaluation on Harmlessness (evaluation set)
Loading...
48.76
Win Rate
SafeDPO
16.0832
24.5666
33.05
41.5334
May 26, 2025
Win Rate
Tie Rate
Lose Rate
Updated 1mo ago
Evaluation Results
Method
Method
Links
Win Rate
Tie Rate
Lose Rate
SafeDPO
Judge Model=GPT-4, Eva...
2025.05
48.76
48.14
3.1
SafeRLHF
Judge Model=GPT-4, Eva...
2025.05
45.98
46.75
7.28
DPO-HARMLESS
Judge Model=GPT-4, Eva...
2025.05
39.94
50.62
9.44
DPO-SAFEBETTER
Judge Model=GPT-4, Eva...
2025.05
26.32
52.32
21.36
DPO-HELPFUL
Judge Model=GPT-4, Eva...
2025.05
17.34
42.72
39.94
Feedback
Search any
task
Search any
task