List your company

Safety Alignment Evaluation on AdvBench (test)

1.6Harm Score (HS)AsFT
1.0244.9128.812.688Jun 10, 2025

Evaluation results

MethodLinks
AsFTUtility Task=AGNEWS2025.061.683.7
Safe LoRAUtility Task=AGNEWS2025.06482.3
SafeInstrUtility Task=AGNEWS2025.064.484.4
Lisa-alignedUtility Task=AGNEWS2025.064.882.6
Lisa-baseUtility Task=AGNEWS2025.067.683.9
SFTUtility Task=AGNEWS2025.0611.283.9
BEAUtility Task=AGNEWS2025.061683.5