ResearchBenchmarksMulti-modal Red-teaming on RTP and PTP (test)Multi-modal Red-teaming on RTP and PTP (test)Follow2.3Refusal Rate (LLaVA)STARE2.1443.1974.255.303May 1, 2026Evaluation resultsMethodMethodLinksRefusal Rate (LLaVA)Refusal Rate (Qwen)Refusal Rate (Gemini)STAREalign=0.2align=0.22026.052.32.95.8Text-Only2026.052.53.14.4Text+SD2026.056.27.810.2