Retrieval Augmented Generation (RAG) enhances LLMs by fusing retrieval-based and generative techniques, addressing LLM limitations like static knowledge, hallucinating, not citing source.