Abstract: The approach of dividing text into chunks and building indexes is a mainstream approach for Retrieval-Augmented Generation (RAG) of Large Language Models (LLMs). However, retrieved text ...