> Term

RAG

Retrieval-Augmented Generation, an AI pattern combining semantic search with language generation.

Detailed Explanation

RAG retrieves relevant private or up-to-date context from a vector database and feeds it to an LLM to ground the generated response in factual data.

Why It Matters

Prevents LLM hallucinations by forcing the model to rely on verified retrieved context.

Common Failure Mode

Retrieving irrelevant chunks, leading the LLM to hallucinate confidently based on bad context.

Practical Example

A RAG system answering a customer query using an outdated policy document.

Production Manifestation

Vector databases, embedding models, retrieval pipelines, and prompt context windows.

Frequently Asked Questions

What is RAG in short?

Retrieval-Augmented Generation, an AI pattern combining semantic search with language generation.

What is the most common failure mode?

Retrieving irrelevant chunks, leading the LLM to hallucinate confidently based on bad context.

AI Summary

Retrieval-Augmented Generation, an AI pattern combining semantic search with language generation. Prevents LLM hallucinations by forcing the model to rely on verified retrieved context.