ResearchBenchmarksSafety Alignment Evaluation on HarmBench (test)Follow14Harm Score (HS)BEA1.524.76811.24Jun 10, 2025Evaluation ResultsMethodMethodLinksHarm Score (HS)False Alarm Rate (FA)BEAUtility Task=AGNEWSUtility Task=AGNEWS2025.061484Lisa-alignedUtility Task=AGNEWSUtility Task=AGNEWS2025.065.884.3SFTUtility Task=AGNEWSUtility Task=AGNEWS2025.065.282.7Lisa-baseUtility Task=AGNEWSUtility Task=AGNEWS2025.064.682.7SafeInstrUtility Task=AGNEWSUtility Task=AGNEWS2025.062.483.2AsFTUtility Task=AGNEWSUtility Task=AGNEWS2025.062.483.4Safe LoRAUtility Task=AGNEWSUtility Task=AGNEWS2025.06281.7