ResearchDatasetsHarmBench and JailbreakBenchFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsAdversarial AttackHarmBench (HB) and JailbreakBench (JBB) (held-out prompts)72.1ASR (%)36