The researchers from Qatar Computing Research Institute share their experiences in building and deploying an application that uses Large Language Models (LLMs) for answering queries over private enterprise documents. They use a framework called Retrieval-Augmented Generation (RAG) to make the application robust and reliable. The system, named Tree-RAG (T-RAG), represents entity hierarchies within the organization in a tree structure. This is used to generate a textual description to augment the context when responding to user queries related to entities within the organization’s hierarchy. The results show that this combination performs better than a simple RAG or finetuning implementation.
Publication date: 12 Feb 2024
Project Page: https://mashable.com/article/samsung-chatgpt-leak-detailsarXiv:2402.07483v1
Paper: https://arxiv.org/pdf/2402.07483
