A technique that enhances generative models by combining them with external information retrieval systems. Instead of relying solely on internal model parameters, RAG retrieves relevant documents or data at inference time. This improves factual accuracy and reduces hallucinations in generated outputs. RAG is widely used in question answering, technical assistants, and knowledge-intensive applications. The approach separates knowledge storage from language generation, increasing scalability and maintainability. In robotics research, RAG is sometimes explored for grounding reasoning in structured knowledge. Overall, RAG represents an important evolution in how intelligent systems access and use information.