arXiv 2605.054092026-07-07cs.AIcs.CL

代理检索增强生成用于财务文档问答

Agentic Retrieval-Augmented Generation for Financial Document Question Answering

College of Computer Science and Technology, Zhejiang University（浙江大学计算机科学与技术学院）

Yang Shu, Yingmin Liu, Zequn Xie

备注：: This paper is withdrawn due to significant methodological errors in the experimental design that fundamentally affect the validity of the results. The errors are not correctable within the current framework, and the conclusions can no longer be supported. We apologize for any inconvenience caused to readers

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出FinAgent-RAG框架，通过迭代检索-推理循环和自我验证，提升金融文档问答的精度。引入对比金融检索器、程序化思维模块和自适应策略路由，实验表明在三个基准数据集上均取得显著效果，准确率提升5.62-9.32个百分点。

摘要：

金融文档问答需要复杂的多步数值推理，涉及异构证据——结构化表格、文本叙述和脚注——分散在企业文件中。现有检索增强生成方法采用单次检索-生成范式，难以处理财务分析中普遍存在的组合推理链。我们提出FinAgent-RAG，一种代理RAG框架，通过迭代检索-推理循环和自我验证，专门针对金融数值推理的精度需求。该框架整合了三个领域特定创新：(1) 通过挖掘困难负样本训练的对比金融检索器，以区分语义相似但数值不同的金融段落；(2) 程序化思维推理模块生成可执行的Python代码进行精确算术运算，而非依赖易出错的LLM基础心理计算；(3) 自适应策略路由器根据问题复杂度动态分配计算资源，减少FinQA上的API成本41.3%，同时保持准确性。在三个基准数据集——FinQA、ConvFinQA和TAT-QA上的广泛实验表明，FinAgent-RAG分别达到76.81%、78.46%和74.96%的执行准确率，优于最强基线5.62-9.32个百分点。消融研究、四种LLM的跨模型评估和部署成本分析证实了该框架的鲁棒性和对金融机构的实际可行性。

英文摘要：

Financial document question answering (QA) demands complex multi-step numerical reasoning over heterogeneous evidence--structured tables, textual narratives, and footnotes--scattered across corporate filings. Existing retrieval-augmented generation (RAG) approaches adopt a single-pass retrieve-then-generate paradigm that struggles with the compositional reasoning chains prevalent in financial analysis. We propose FinAgent-RAG, an agentic RAG framework that orchestrates iterative retrieval-reasoning loops with self-verification, specifically engineered for the precision requirements of financial numerical reasoning. The framework integrates three domain-specific innovations: (1) a Contrastive Financial Retriever trained with hard negative mining to distinguish semantically similar but numerically distinct financial passages, (2) a Program-of-Thought reasoning module that generates executable Python code for precise arithmetic rather than relying on error-prone LLM-based mental computation, and (3) an Adaptive Strategy Router that dynamically allocates computational resources based on question complexity, reducing API costs by 41.3% on FinQA while preserving accuracy. Extensive experiments on three benchmark datasets--FinQA, ConvFinQA, and TAT-QA--demonstrate that FinAgent-RAG achieves 76.81%, 78.46%, and 74.96% execution accuracy respectively, outperforming the strongest baseline by 5.62--9.32 percentage points. Ablation studies, cross-backbone evaluation with four LLMs, and deployment cost analysis confirm the framework's robustness and practical viability for financial institutions.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。