ResearchBenchmarksSafety Content Detection on Standard non-adversarial dataset RacismFollow99TPRGAVEL96.9297.469898.54Jan 27, 2026Evaluation ResultsMethodMethodLinksTPRFPRGAVEL2026.01992GPT-4Rule Definitions=NoneRule Definitions=None2026.01990GPT-4Rule Definitions=IncludedRule Definitions=Included2026.01975