List your company

Safety Alignment Evaluation on HarmBench (test)

14Harm Score (HS)BEA
1.524.76811.24Jun 10, 2025

Evaluation results

MethodLinks
BEAUtility Task=AGNEWS2025.061484
Lisa-alignedUtility Task=AGNEWS2025.065.884.3
SFTUtility Task=AGNEWS2025.065.282.7
Lisa-baseUtility Task=AGNEWS2025.064.682.7
SafeInstrUtility Task=AGNEWS2025.062.483.2
AsFTUtility Task=AGNEWS2025.062.483.4
Safe LoRAUtility Task=AGNEWS2025.06281.7