ResearchDatasetsHuman-LabeledFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsSafety DetectionHuman-labeled evaluation set92.86Unsafe Recall (HarmBench)4Forced AlignmentHuman-Labeled (test)0.007Avg. RTF4