1
Dynamic Optimization and Safety Indicator Injection for Jailbreaking Text-to-Image Models with Multimodal Safety Filters
最新研究揭示：通过动态优化与安全指示注入，可破解T2I模型多模态安全过滤器，LLM也无法阻挡。
arXiv:2505.18979v2 Announce Type: replace Abstract: Text-to-image (T2I) models can generate not-safe-for-work (NSFW) content, motivating multi-stage s…