A Dataset of Robot-Patient and Doctor-Patient Medical Dialogues for Spoken Language Processing Tasks

Researchers released MeDial-Speech, a free non-commercial speech dataset of realistic medical consultations.

The 111+ hour dataset spans robot-patient and doctor-patient dialogues across four conditions: Lewy body dementia, heart failure, shoulder pain, and angina.

The team also tested GPT-5 mini, DeepSeek-V3, and Claude Sonnet 4 on a sentence-selection benchmark; Claude Sonnet 4 performed best.

All three models showed strong overconfidence, underscoring that current LLMs are still unreliable for medical dialogue understanding.