FinTMMBench
收藏资源简介:
FinTMMBench是一个综合性的金融分析数据集,由新加坡国立大学等多家机构合作构建。该数据集包含了来自纳斯达克100家公司的财务表格、新闻文章、每日股价以及视觉技术图表等多种模态的数据。数据集通过模板引导的方式自动生成问题回答对(QA对),涵盖了信息提取、趋势分析、情感分析等10种金融分析任务,能够全面评估时序感知多模态RAG系统在金融领域中的应用能力。
FinTMMBench is a comprehensive financial analysis dataset co-constructed by multiple institutions including the National University of Singapore. This dataset encompasses multi-modal data from 100 NASDAQ-listed companies, such as financial statements, news articles, daily stock prices, and visual technical charts. It automatically generates question-answer (QA) pairs via template-guided methods, covering 10 financial analysis tasks including information extraction, trend analysis, and sentiment analysis. This benchmark can comprehensively evaluate the application capabilities of temporal-aware multimodal RAG systems in the financial domain.

- 1FinTMMBench: Benchmarking Temporal-Aware Multi-Modal RAG in Finance新加坡国立大学 · 2025年



