ResearchTasksModel Security EvaluationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast Updated3,312 safety test cases (derived from 81 baseline prompts) (overall)Claude Sonnet 437.5Refusal Rate3Feb 26, 2026
3,312 safety test cases (derived from 81 baseline prompts) (overall)Claude Sonnet 437.5Refusal Rate3Feb 26, 2026