FinRAGBench-V
收藏资源简介:
FinRAGBench-V是一个为金融领域定制的多模态RAG基准,有效整合了多模态数据并提供视觉引用以确保可追溯性。它包括一个包含60,780个中文页面和51,219个英文页面的双语检索语料库,以及一个高质量、人工标注的问答(QA)数据集,跨越异构数据类型和七个问题类别。此外,我们引入了RGenCite,一个RAG基线,它无缝地集成了视觉引用和生成。此外,我们提出了一种自动引用评估方法,系统地评估多模态大型语言模型(MLLMs)的视觉引用能力。在RGenCite上进行的广泛实验强调了FinRAGBench-V的挑战性,为金融领域多模态RAG系统的发展提供了宝贵的见解。
FinRAGBench-V is a multimodal RAG benchmark tailored for the financial domain, which effectively integrates multimodal data and provides visual citations to ensure traceability. It comprises a bilingual retrieval corpus containing 60,780 Chinese pages and 51,219 English pages, as well as a high-quality, manually annotated question answering (QA) dataset spanning heterogeneous data types and seven question categories. Furthermore, we introduce RGenCite, a RAG baseline that seamlessly integrates visual citations and generation. Additionally, we propose an automatic citation evaluation method to systematically assess the visual citation capabilities of multimodal large language models (MLLMs). Extensive experiments conducted on RGenCite highlight the challenging nature of FinRAGBench-V and provide valuable insights for the development of multimodal RAG systems in the financial domain.




