ResearchDatasetsHarmBench and DEFCONFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsMulti-turn Jailbreak DetectionHarmBench and DEFCON Multi-turn Jailbreak N=1,010 (test)84F1 Score10