AmazonScience/GaRAGe

GaRAGe是一个用于评估RAG（检索增强生成）的大型基准数据集，包含人工审核的长篇答案和每个接地段落的注释。它包含2366个不同复杂性、动态性和主题的问题，以及从私人文档集和网络中检索的超过35K个注释段落。这个数据集适用于评估LLM在生成RAG答案时识别相关信息的性能。

GaRAGe is a large benchmark for RAG (Retrieval-Augmented Generation) evaluation with human-curated long-form answers and annotations for each grounding passage. It contains 2366 questions of diverse complexity, dynamism, and topics, along with over 35K annotated passages retrieved from both private document sets and the Web. This dataset is suitable for evaluating the performance of LLMs in identifying relevant information when generating RAG answers.