Answer

Retrieval-Augmented Generation: retrieve relevant documents from a knowledge base and inject them into the LLM context. Reduces hallucination by grounding answers in real data. Enables real-time knowledge updates without retraining the model.