ResearchBenchmarksSafety Alignment Evaluation on BeaveTails (test)Follow37.2HS ScoreSFT13.48819.64425.831.956Jun 10, 2025Evaluation ResultsMethodMethodLinksHS ScoreFA ScoreSFTUtility Task=AGNEWSUtility Task=AGNEWS2025.0637.284.9BEAUtility Task=AGNEWSUtility Task=AGNEWS2025.0636.884.2Lisa-alignedUtility Task=AGNEWSUtility Task=AGNEWS2025.0631.485.8Lisa-baseUtility Task=AGNEWSUtility Task=AGNEWS2025.0630.883.1SafeInstrUtility Task=AGNEWSUtility Task=AGNEWS2025.0621.683.2Safe LoRAUtility Task=AGNEWSUtility Task=AGNEWS2025.0618.882.6AsFTUtility Task=AGNEWSUtility Task=AGNEWS2025.0614.482.9