ResearchTasksAdversarial Attack DefenseFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedGCG IndividualPerplexity Defense100BAR18Feb 26, 2026Safety Evaluation Set GCG, AutoDAN, DeepInception, Prefilling, Intent Laundering n=200 (full public set)AlphaSteer0GCG Attack Success Rate11Jun 30, 2026RedTeamCUADEACTION0.02ASR10Feb 26, 2026GCG-TransferVicuna-7B-chat-HF99.3BAR4Feb 26, 2026Held-out attacks (test)SecureCAI7.8ASR (Multi-turn Manip.)2Feb 26, 2026
Safety Evaluation Set GCG, AutoDAN, DeepInception, Prefilling, Intent Laundering n=200 (full public set)AlphaSteer0GCG Attack Success Rate11Jun 30, 2026