Acronyms Dictionary
RAG
Definition
Retrieval-Augmented Generation. A technique for enhancing LLM output with outside data.
Deep Dive
RAG is a framework for improving the accuracy and reliability of LLM responses. It works by first "retrieving" relevant facts from an external knowledge base (like a company's internal docs) and then "augmenting" the LLM's prompt with that data.
Examples & Use Cases
- 1An AI support agent using RAG to pull current shipping data before answering a user
- 2Reducing "Hallucinations" in an LLM by grounding its answers in a verified database via RAG
- 3Implementing RAG to allow an AI to "remember" and reference a user's specific history
Related Terms
LLMVector DatabaseSemantic Search