# Learning to Read the Contextual Tokens in Diffusion Transformers > Source: > Published: 2026-10-07 09:00:42+00:00 Multimodal Diffusion Transformers (MM-DiTs) jointly process visual and textual representations throughout generation. These models repeatedly update the text tokens through multimodal attention, forming dynamic contextual tokens whose function is not well understood. In this work, we introduce a fra