Path-Constrained Retrieval: A Structural Approach to Reliable LLM Agent Reasoning Through Graph-Scoped Semantic Search

TL;DR

This work tackles the problem of incoherent LLM reasoning when retrieved context lacks alignment with the agent's reasoning state. It introduces Path-Constrained Retrieval (PCR), which confines retrieval to nodes reachable from a chosen anchor in a knowledge graph and combines this with semantic ranking. On PathRAG-6, PCR achieves perfect structural consistency across six domains while maintaining competitive relevance, and substantially reduces the graph distance between anchor and retrieved nodes, demonstrating improved reliability and coherence in reasoning. The approach offers practical benefits for LLM agents by better aligning retrieved information with the reasoning context, with acceptable computational overhead and clear paths for future enhancements.

Abstract

Large Language Model agents often retrieve context from knowledge bases that lack structural consistency with the agent's current reasoning state, leading to incoherent reasoning chains. We introduce Path-Constrained Retrieval (PCR), a retrieval method that combines structural graph constraints with semantic search to ensure retrieved information maintains logical relationships within a knowledge graph. PCR restricts the search space to nodes reachable from an anchor node, preventing retrieval of structurally disconnected information that may lead to inconsistent reasoning. We evaluate PCR on PathRAG-6, a benchmark spanning six domains with 180 nodes and 360 edges. Our results show that PCR achieves full structural consistency compared to 24-32 percent in baseline methods, while maintaining strong relevance scores. On the technology domain, PCR obtains full relevance at rank 10 with full structural consistency, significantly outperforming vector search and hybrid retrieval. PCR reduces the average graph distance of retrieved context by 78 percent compared to baselines, demonstrating retrieval of more structurally consistent information. These findings suggest that path-constrained retrieval is an effective approach for improving the reliability and coherence of LLM agent reasoning systems.