ResearchBenchmarksSafety Alignment Evaluation on Harmful (test)Follow4Harmful Score (HS)AsFT3.4567.12810.814.472Jun 10, 2025Evaluation ResultsMethodMethodLinksHarmful Score (HS)False Acceptance Rate (FA)AsFTUtility Task=AGNEWSUtility Task=AGNEWS2025.06484.3Safe LoRAUtility Task=AGNEWSUtility Task=AGNEWS2025.065.681.2BEAUtility Task=AGNEWSUtility Task=AGNEWS2025.0616.484.4Lisa-alignedUtility Task=AGNEWSUtility Task=AGNEWS2025.0616.881.8SafeInstrUtility Task=AGNEWSUtility Task=AGNEWS2025.0616.884.3Lisa-baseUtility Task=AGNEWSUtility Task=AGNEWS2025.0617.273.5SFTUtility Task=AGNEWSUtility Task=AGNEWS2025.0617.684.3