Unified Safe In-context Image Generation in Multimodal Diffusion Transformers via Restricting Unsafe Information Flows 事件

Unified Safe In-context Image Generation in Multimodal Diffusion Transformers via Restricting Unsafe Information Flows arXiv:2606.06875v1 Announce Type: new Abstract: Diffusion transformers (DiTs) equipped with multimodal attention (MM-Attn) have become a dominant paradigm for image generation. However, preventing the generation of harmful content remains a critical challenge, particularly in image-to-image (I2I) editing tasks. Existing safety mechanisms are primarily designed for text-to-image