The abstract is read at the publisher; the summary is JClub's.
Medical retrieval-augmented generation (RAG) is still at an early stage, with research concentrated on English-centric retrieval and generic LLMs, and evaluation that underemphasizes bias and safety.
- Why it matters: Rapidly expanding medical knowledge and complex clinical workflows create a need to improve how language models support clinical tasks despite their inherent limitations.
- What they did: This scoping review synthesizes RAG implementations in medicine, noting reliance on publicly available data, English-centric embedding models for retrieval, mostly generic LLMs, and evaluation using both automated metrics and human judgments.
- The result: The review finds RAG is mainly applied to question answering, report generation, text summarization, and information extraction, and it highlights needs for clinical validation, cross-linguistic adaptation, and low-resource support for trustworthy global use.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.