Discrete Diffusion Language Models for Interactive Radiology Report Drafting
This paper swaps standard left‑to‑right text models for diffusion‑style text generators in radiology. The model drafts reports faster, lets doctors edit anywhere in the text, and still matches or beats strong baselines. It is a concrete pattern for building medical drafting tools that feel more like collaborative editing than one‑shot replies.
Max Van Puyvelde, Halil Ibrahim Gulluk