Clinician use of language models diverges from how the models are evaluated Paper • 2610.11069 • Published Oct 8 • 4