MultiHop-RAG
收藏资源简介:
MultiHop-RAG是由香港科技大学开发的一个专注于多跳查询的RAG(Retrieval-Augmented Generation)数据集。该数据集包含一个知识库、大量多跳查询、它们的正确答案以及相关的支持证据。数据集的构建过程利用了英语新闻文章数据集作为基础知识库。MultiHop-RAG旨在解决现有RAG系统在处理需要检索和推理多条支持证据的多跳查询时的不足。数据集的应用领域包括推理、比较和时序查询,旨在评估和提升大型语言模型在实际应用中的性能。
MultiHop-RAG is a Retrieval-Augmented Generation (RAG) dataset focused on multi-hop queries, developed by The Hong Kong University of Science and Technology. This dataset includes a knowledge base, a large number of multi-hop queries, their correct answers, and relevant supporting evidence. The dataset was constructed using English news article datasets as the underlying knowledge base. MultiHop-RAG aims to address the shortcomings of existing RAG systems when handling multi-hop queries that require retrieving and reasoning over multiple supporting evidence pieces. Its application areas cover reasoning, comparative, and temporal queries, and it is designed to evaluate and improve the performance of large language models (LLMs) in real-world applications.




