EHRNote-ChatQA: A Benchmark for Evidence-Grounded Multi-Turn Clinical Question Answering over Longitudinal Discharge Summaries
EHRNoteChatQA is a benchmark for evaluating AI's ability to answer complex clinical questions across multiple patient discharge summaries. It tests if models can provide sequential answers while accurately grounding every response in specific, verifiable evidence scattered across several longitudinal patient notes. Findings show current models struggle with this multi-turn evidence grounding.