ResearchTasksAdversarial Prompt Injection AttackFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast Updated1000 images Gemini-2.5 (hard criterion)CoTTA73.2ASR2Apr 1, 20261000 images GPT-5 hard criterionFOA-Attack22.8ASR2Apr 1, 20261000 images GPT-4o hard criterionCoTTA64.7Attack Success Rate (ASR)2Apr 1, 2026