PhysAssistBench: Benchmark for LLMs in physician assistance

Arxiv CS.CL (NLP/language models)

The benchmark evaluates LLMs on interactive doctor-patient-EHR tasks, testing multi-turn conversation, clinical reasoning, and EHR manipulation. Current models show significant gaps in integrating EHR context and handling complex patient interactions.

LLM

Arxiv CS.CL (NLP/language models)·Tianming DuPeijie YuSihan ShangDanli ShiMy Linh NguyenShengbo GaoGuangyuan LiYinghong YuYan JiangQianlong ZhaoBehzad BozorgtabarShaoxiong JiJiazhen PanDaniel RueckertJiancheng Yang

Jun 18, 4:00 AM