AuditAgent: Expert-Guided Multi-Agent Reasoning for Cross-Document Fraudulent Evidence Discovery

2025-09-30

AI Summary

Financial fraud evidence is often fragmented and concealed across multi-year financial disclosure documents, posing significant challenges for cross-document, fine-grained evidence localization. To address this, we propose a multi-agent reasoning framework infused with audit-domain expertise, featuring three key innovations: (1) entity-level risk prior modeling, (2) a hybrid retrieval strategy integrating semantic and structural cues, and (3) a specialized agent division-of-labor mechanism for interpretable evidence chain discovery and aggregation. Our method leverages expert-annotated datasets and integrates domain-enhanced collaborative reasoning, cross-document retrieval, and evidence fusion techniques. Evaluated in real-world regulatory settings, it achieves substantially higher evidence recall than generic multi-agent baselines, while ensuring transparent, auditable, and controllable reasoning. This work establishes the first automated, explainable, and domain-customized reasoning paradigm specifically designed for financial forensics.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Data Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity

Abstract

Financial fraud detection in real-world scenarios presents significant challenges due to the subtlety and dispersion of evidence across complex, multi-year financial disclosures. In this work, we introduce a novel multi-agent reasoning framework AuditAgent, enhanced with auditing domain expertise, for fine-grained evidence chain localization in financial fraud cases. Leveraging an expert-annotated dataset constructed from enforcement documents and financial reports released by the China Securities Regulatory Commission, our approach integrates subject-level risk priors, a hybrid retrieval strategy, and specialized agent modules to efficiently identify and aggregate cross-report evidence. Extensive experiments demonstrate that our method substantially outperforms General-Purpose Agent paradigm in both recall and interpretability, establishing a new benchmark for automated, transparent financial forensics. Our results highlight the value of domain-specific reasoning and dataset construction for advancing robust financial fraud detection in practical, real-world regulatory applications.

Problem

• Detects subtle financial fraud evidence across documents
• Localizes fine-grained evidence chains using multi-agent reasoning
• Addresses evidence dispersion in complex multi-year financial disclosures

Innovation

• Multi-agent reasoning framework with auditing expertise
• Hybrid retrieval strategy for cross-report evidence aggregation
• Domain-specific dataset integration for enhanced fraud detection