ResearchDatasetsWildJailbreak × StrongREJECTWildJailbreak × StrongREJECTFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsDirectional Jailbreak DetectionWildJailbreak (Safe) × StrongREJECT (Harmful)98.1AUROC (Mono)3