A Dataset for Evaluating Retrieval-Augmented Generation Across Documents
Dataset Description
MultiHop-RAG: a QA dataset to evaluate retrieval and reasoning across documents with metadata in the RAG pipelines. It contains 2556 queries, with evidence for each query distributed across 2 to 4 documents. The queries also involve document metadata, reflecting complex scenarios commonly found in real-world RAG applications.
Dataset Sources… See the full description on the dataset page: https://huggingface.co/datasets/yixuantt/MultiHopRAG.
