SynthBench

该数据集是一个涵盖4个领域内8个特定领域文档的基准，其中包含了不同查询复杂性、线索完整性和细粒度的引用粒度。该数据集旨在优化检索器的健壮性和生成器的准确性，已经证明它能够提高不同领域内RAG系统的性能。这是一个大规模的合成数据集，包含了单跳和多跳查询，用于对RAG系统进行基准测试。

This dataset is a benchmark encompassing 8 domain-specific documents across 4 distinct domains, which incorporates varying query complexities, different levels of clue completeness, and fine-grained reference granularities. It aims to optimize the robustness of retrievers and the accuracy of generators, and has been proven to improve the performance of Retrieval-Augmented Generation (RAG) systems across various domains. This is a large-scale synthetic dataset containing both single-hop and multi-hop queries, specifically designed for benchmarking RAG systems.