How Retrieval-Augmented Generation (RAG) Extends LLM Knowledge Beyond Training Data
English summary
This tutorial explains that language model knowledge is frozen after training, and introduces Retrieval-Augmented Generation (RAG) as a method to let LLMs read new information such as private documents or real-time data. It highlights RAG’s role in giving models access to up-to-date answers beyond their original training cut-off.
Chinese summary
该教程解释了大语言模型的知识在训练结束后便固定下来，并介绍了检索增强生成（RAG）方法，使模型能访问私有文档或实时数据等新信息。文章强调了RAG在帮助模型提供超出原始训练时间线的最新答案中的作用。
Key points
LLM knowledge is static after training concludes.
LLM的知识在训练结束后是静态的。
RAG enables LLMs to access and integrate information not in their training data.
RAG使LLM能够访问并整合训练数据中不包含的信息。
RAG can use private documents or recent information to provide up-to-date answers.
RAG可以使用私有文档或最新信息来提供实时的答案。