Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Toxicity Steering on Eso-LM (512 sequences)

-9.2Toxicity Score

DDPO

-9.54-7.245-4.95-2.655Sep 25, 2025
Updated 1d ago

Evaluation Results

MethodLinks
2025.09
-9.2
2025.09
-9.2
2025.09
-9.2
2025.09
-9.1
2025.09
-8.9
2025.09
-8.9
2025.09
-8.6
2025.09
-8.5
2025.09
-7.3
2025.09
-5.5
2025.09
-2.7
2025.09
-0.7