ResearchTasksFinetuning attack successFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedHarmful HEx-PHI (adversarial)SafeLM84.92ASR9Jun 18, 2026GSM8K (benign)SafeLM56ASR9Jun 18, 2026