ResearchTasksPrompt Injection Attack MitigationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedSkill-Inject 139 sandboxes (full set)Deepseek-V4-Flash+OC2.9ASR11Jun 2, 2026