Citation-Closure Retrieval and Per-Rule Attribution for Real-World Regulatory Compliance Question Answering

2026-05-28

AI Summary

Current RAG systems struggle with multi-level regulatory citation in compliance question answering, suffering from flattened referencing, fragmented retrieval, and fragile attribution. This work formalizes this task for the first time and introduces RefWalk, a framework that traverses cross-document citation closures via a knowledge graph, integrates multi-perspective candidate answers through shared thematic anchors, and enables explicit per-rule attribution. The authors also construct RegOps-Bench, a new benchmark featuring structured regulatory modeling and contrastive evaluation protocols. Experimental results demonstrate that RefWalk substantially improves retrieval recall and citation accuracy, while uncovering a critical performance bottleneck in existing systems caused by their reliance on flat rule representations.

Abstract

Deploying Large Language Models (LLMs) for regulatory compliance demands rigorous traceability via comprehensive citations across multi-tiered authority structures. Unlike traditional multi-hop or legal QA, this task requires structured procedural lookups and evidence-set closure rather than entity resolution or case-law reasoning. Existing RAG systems struggle here due to flattened citation edges, fragmented retrieval expansions, and fragile post-hoc attribution. We formalize Regulatory Compliance QA with RegOps-Bench, a novel benchmark featuring an Operational Knowledge Graph derived from complex national R\&D regulations. To address these bottlenecks, we propose RefWalk, a unified framework driven by a shared topic anchor. RefWalk traverses cross-document citations, fuses multi-view candidates via max-based aggregation, and enforces per-rule attribution to explicitly map claims to sources. We establish a strong baseline with substantial improvements in retrieval recall and citation accuracy. Finally, a contrastive evaluation on a U.S. health compliance dataset (HIPAA) reveals that existing systems exhibit saturation on flat-structure rules, underscoring the need for RegOps-Bench. Our code is available at https://github.com/yeongjoonJu/RefWalk.

Problem

• Regulatory Compliance QA
• Citation Closure
• Per-Rule Attribution
• Evidence-Set Closure
• Multi-tiered Authority

Innovation

• Regulatory Compliance QA
• Citation-Closure Retrieval
• Per-Rule Attribution
• Knowledge Graph
• RefWalk