Agentic Retrieval-Augmented Generation for Financial Document Question Answering

TL;DR

The paper introduces FinAgent-RAG, an innovative retrieval-augmented generation framework for financial QA that enhances multi-step numerical reasoning through iterative retrieval, programmatic reasoning, and adaptive resource allocation.

Contribution

It presents a novel agentic RAG approach with domain-specific retrieval, code-based reasoning, and dynamic strategy routing, improving accuracy and reducing costs in financial document question answering.

Findings

Achieves over 74% accuracy on benchmark datasets.

Reduces API costs by 41.3% with adaptive strategies.

Outperforms existing baselines by 5.62--9.32 percentage points.

Abstract

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.