RAG
Retrieval-Augmented Generation — a technique that grounds AI responses in retrieved factual documents to reduce hallucinations.
Full Definition
Retrieval-Augmented Generation (RAG) is an AI architecture pattern that combines information retrieval with language model generation. When an AI agent receives a query, the RAG system first searches a knowledge base (typically using vector embeddings) to find relevant documents, then provides these documents as context to the language model for generating its response. RAG significantly reduces hallucinations by grounding the model's outputs in factual, verified source material rather than relying solely on the model's parametric knowledge. However, RAG is not a complete solution — the model can still misinterpret retrieved documents, the knowledge base may contain outdated information, and retrieval failures can lead to irrelevant context. RAG is most effective when combined with additional governance measures like semantic consistency checking and citation verification.
Related Terms
AI Hallucination
When an AI model generates information that appears plausible but is factually incorrect, fabricated, or unsupported by its input data.
AI Agent
An autonomous software system that uses AI models to perceive its environment, make decisions, and take actions to achieve goals.
Anomaly Detection
The automated identification of unusual patterns or behaviors in AI agent operations that deviate from expected norms.