ResearchTasksText ModerationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedHarmBench n = 400Gemma 3-1B42Flagged Count13Apr 30, 2026WildJailbreak Adv. Benign n = 210AdaSteer3Flagged Count13Apr 30, 2026WildJailbreak Adv. Harmful n = 2,000Gemma 3-1B278Flagged Count13Apr 30, 2026