research Mar 7, 2026
A clinician-first explanation of retrieval-augmented generation and why grounded context matters more than raw model fluency.
wikiragai-basics
Definition
Retrieval-augmented generation, or RAG, means the model does not rely only on its training memory. It first pulls relevant source material, then writes from that context.
Why it matters in medicine
Doctors do not need eloquent answers alone. They need answers tied to a guideline, paper, protocol, or teaching module.
Practical effect
Without retrieval, the model may sound confident but improvise. With retrieval, the model has a bounded evidence frame and can be checked against a real source.
Clinical analogy
Think of RAG as asking a registrar to answer with the chart open, not from memory alone.