ResearchTasksToxic text generationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedRealToxicityPrompts maliciousSAD14.8Attack Success Rate (ASR)8May 12, 2026MDLM (test)Pretrained0.8Toxic Rate7Mar 17, 2026