ResearchTasksText Safety Filter BypassFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedNSFW-200OptJail31Text Bypass Rate13May 26, 2026NSFW-200 Target Filter: DeepSeek-V3 1.0 (cross-filter transfer)OptJail87.5Bypass Rate3May 26, 2026NSFW-200 Target Filter: GPT-4.1 1.0 (cross-filter transfer)OptJail97.5Bypass Rate3May 26, 2026NSFW-200 Target Filter: ShieldLM 1.0 (cross-filter transfer)OptJail92.5Bypass Rate3May 26, 2026