ResearchTasksToxicity DefenseFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedLLaVA v1.5NullSteer22.35Toxicity Score36Mar 24, 2026Qwen2-VLNullSteer0.05Toxicity Score36Mar 24, 2026MiniGPT-4NullSteer2.89Toxicity Score36Mar 24, 2026PGD-ToxicASTRA1.53ASR (PGD-Toxic)14Mar 31, 2026