Retrieval-Augmented Generation (RAG)
noun

Definition:
A generative AI framework where a large language model retrieves relevant external data—such as documents, web pages, or knowledge base entries—in real time to inform and enhance its generated responses. This approach improves factual accuracy and reduces hallucination by grounding answers in up-to-date or domain-specific sources.

Usage:
“Perplexity’s search engine is based on retrieval-augmented generation to ensure citations in its answers are verifiable.”

Compare:
Open-book Question Answering, Hybrid AI, Generative AI