ResearchTasksHarmful score evaluationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedBeaverTails (test)SFT28.4Harmful Score52May 15, 2026SorrybenchSurgery12.95Harmful Score8Feb 26, 2026