ResearchTasksLanguage model detoxificationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedRealToxicityPrompts (test)TruthFlow91.3Distinct-154Feb 26, 2026Human Evaluation 50 generations (test)LM-Steer0.49Detoxification Count6Feb 26, 2026