ResearchTasksFinetuning attack successFinetuning attack successFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedHarmful HEx-PHI (adversarial)SafeLM84.92ASR9Jun 18, 2026GSM8K (benign)SafeLM56ASR9Jun 18, 2026