ResearchTasksEnd-to-End Attack EffectivenessFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedPoisonedSkills 1,070 adversarial samplesSonnet 4.6919Refusal Count8Apr 6, 2026Skill corpusVulMask58.8ASR2Jun 18, 2026