Tag
This paper introduces MeDial-Speech, a dataset of robot-patient and doctor-patient medical dialogues for spoken language processing, and evaluates three LLMs on a sentence selection benchmark, finding Claude Sonnet 4 most accurate.