The survey discusses Retrieval-Augmented Generation (RAG) which integrates up-to-date external knowledge into Large Language Models (LLMs), enhancing their capabilities in generating high-quality outputs. The study covers foundational knowledge on LLMs and dives into the architectures, training strategies, and applications of RAG-enhanced LLMs. - Comprehensive review: The paper provides a detailed review of recent studies on RAG within LLMs, highlighting the technical aspects like architectures, training strategies, and applications. - Functional benefits: Demonstrates how RAG enables LLMs to incorporate up-to-date knowledge, offering more accurate and contextually relevant outputs. - Challenges tackled: The survey addresses challenges like hallucinations and outdated internal knowledge in LLMs. - Real-world applications: Discusses how RAG can be effectively applied across various sectors including healthcare and customer service. - Future directions: It identifies gaps in the current literature and suggests several promising directions for future research. Opinion: This paper is crucial for understanding how RAG can transform LLMs by providing them with the capability to access reliable external knowledge sources. Its extensive review and analysis set the stage for future innovations in AI technologies.