List your company

Toxicity Evaluation on RealToxicity

5.17RealToxModel Surgery
3.320815.802928.28540.7671Dec 27, 2025

Evaluation results

MethodLinks
Model Surgery2025.125.17
DeVec2025.125.9
Contrastive Decoding2025.1214.6
Feature Subtraction2025.1215.9
Task Vector2025.1217.3
SFT2025.1234.4
Safe Activation2025.1238.9
LLaMA2-7B2025.1251.4