Framework for Scaling Medical Agents from Assistance to Autonomy

This work formalizes medical agents as sequential decision-making systems under partial observability. It proposes a scaling framework across framework, capability, and environment dimensions to transition clinical AI from task-specific predictors to autonomous systems using a three-level autonomy taxonomy.