A two‐stage active learning pipeline that uses diffusion‐based feature sampling and entropy‐augmented disagreement to pick the most informative pixels under extreme labeling constraints.

We introduce PatientSim, a patient simulator that generates realistic and diverse patient personas for clinical scenarios.

The paper introduces a novel latent diffusion model that uses local vicinity structures to achieve state-of-the-art domain adaptation, preserving privacy without needing source data during adaptation.