RAG (Retrieval-Augmented Generation)

What is RAG (Retrieval-Augmented Generation)?

RAG reduces hallucination and gives users source citations. Quality depends on ingestion, chunking, embeddings, ranking, and freshness — “upload documents and chat” is the demo, not the architecture. See RAG Technology Best Practices.

What is an example of RAG (Retrieval-Augmented Generation)?

When an employee asks about travel policy, the system retrieves the current policy clauses, passes only the relevant passages to the model, and answers with links to the exact source sections.

How is RAG (Retrieval-Augmented Generation) different from related concepts?

RAG adds knowledge at inference time without changing model weights. Fine-tuning changes model behavior or knowledge encoded in weights. Search returns sources; RAG uses retrieved sources to generate an answer.

Why it matters for on-premise & regulated AI

RAG determines what your AI knows — and therefore what it can leak. Enterprises deploying RAG over regulated data need the full pipeline (chunking, embedding, indexing, retrieval) to run on infrastructure they control, with document-level permissions enforced at query time. That is the difference between RAG as a demo and RAG as a system a CISO signs off on.

What should enterprises evaluate for RAG (Retrieval-Augmented Generation)?

• Measure retrieval recall, ranking quality, answer correctness, claim support, citation accuracy, and abstention separately.
• Test with real documents, languages, permissions, table-heavy content, ambiguous queries, and questions with no answer.
• Track source freshness and make every citation resolvable to the version used during generation.

Related terms

• Private RAG
• On-Premises RAG
• Self-Hosted RAG

Authoritative sources

Primary sources for the formal meaning, requirements, or original research behind RAG (Retrieval-Augmented Generation):

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks — source for RAG (Retrieval-Augmented Generation) NeurIPS