ResearchBenchmarksSafety Content Detection on Standard non-adversarial dataset RomanceSafety Content Detection on Standard non-adversarial dataset RomanceFollow100TPRGAVEL94.896.1597.598.85Jan 27, 2026Evaluation resultsMethodMethodLinksTPRFPRGAVEL2026.011001GPT-4Rule Definitions=NoneRule Definitions=None2026.011000GPT-4Rule Definitions=IncludedRule Definitions=Included2026.01950