RAG

RAG

A technique where a system retrieves relevant documents before answering a question, then generates its response grounded in that retrieved content — letting a model draw on information it wasn't trained on, like recent or internal documents.

Briefings mentioning this
2026-08-17