REGREACT: Self-Correcting Multi-Agent Pipelines for Structured Regulatory Information Extraction

2026-04-13

AI Summary

This work addresses the challenges of extracting structured compliance requirements from regulatory documents, including model hallucination, loss of hierarchical relationships, and difficulties in resolving cross-document dependencies. The authors propose REGREACT, a novel seven-stage self-correcting multi-agent framework in which each stage employs an Observe–Diagnose–Repair (ODR) loop to iteratively refine outputs. By integrating external legal document retrieval and summary embeddings, the framework generates self-contained, structured representations that not only mitigate hallucinations but also detect and correct erroneous cross-references within regulations, constructing a typed criterion graph to preserve structural fidelity. Evaluated on a newly curated dataset comprising 242 activities and over 4,800 hierarchical criteria derived from three EU Delegated Acts, the approach significantly outperforms a GPT-4o single-pass baseline across all structural and semantic metrics.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Search and Optimization: Metareasoning and Metaheuristics
• Multiagent Systems: Modeling other Agents

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
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Extracting structured, machine-readable compliance criteria from regulatory documents remains an open challenge. Single-pass language models hallucinate structural elements, lose hierarchical relationships, and fail to resolve inter-document dependencies. We introduce \textsc{RegReAct}, a self-correcting multi-agent framework that decomposes regulatory information extraction into seven specialized stages, each with an \textit{Observe--Diagnose--Repair} (ODR) loop that validates outputs against the source, correcting not only model hallucinations but also cross-reference errors in the regulations themselves. To ensure structural accuracy, \textsc{RegReAct} constructs a typed criterion graph; to ensure completeness, it resolves external dependencies by retrieving, summarizing, and embedding referenced legal content inline, producing self-contained outputs. Applying \textsc{RegReAct} to three EU Taxonomy Delegated Acts, we construct a dataset comprising 242 activities with over 4,800 hierarchical criteria, thresholds, and enriched source summaries. Evaluation against a GPT-4o single-pass baseline confirms that \textsc{RegReAct} outperforms it across all structural and semantic metrics. Code and data will be made publicly available: https://github.com/RECOR-Benchmark/RECOR

Problem

• regulatory information extraction
• structured compliance criteria
• multi-agent systems
• hallucination
• inter-document dependencies

Innovation

• multi-agent framework
• self-correcting
• structured information extraction
• regulatory compliance
• cross-reference resolution