ResearchTasksSkill-based attack evaluationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedSkillHarm Self-Mutating PoisoningGPT-5.469.3ASR6Jun 2, 2026SkillHarm Fixed-Payload PoisoningGPT-5.486.3ASR6Jun 2, 2026