KnowPath: Knowledge-enhanced Reasoning via LLM-generated Inference Paths over Knowledge Graphs

TL;DR

KnowPath addresses LLM hallucinations in knowledge-based QA by tightly integrating the model's internal knowledge with external knowledge graphs. It introduces a three-stage framework: Inference Paths Generation, Subgraph Exploration, and Evaluation-based Answering, enabling interpretable, directed KG reasoning guided by the LLM's own reasoning. Empirical results across four KGQA datasets show state-of-the-art accuracy and notable cost-efficiency, with clear gains on multi-hop tasks and open-domain questions. The work emphasizes interpretability through explicit subgraphs and reasoning paths, offering a practical, plug-and-play approach for robust factual QA.

Abstract

Large language models (LLMs) have demonstrated remarkable capabilities in various complex tasks, yet they still suffer from hallucinations. By incorporating and exploring external knowledge, such as knowledge graphs(KGs), LLM's ability to provide factual answers has been enhanced. This approach carries significant practical implications. However, existing methods suffer from three key limitations: insufficient mining of LLMs' internal knowledge, constrained generation of interpretable reasoning paths, and unclear fusion of internal and external knowledge. Therefore, we propose KnowPath, a knowledge-enhanced large model framework driven by the collaboration of internal and external knowledge. It relies on the internal knowledge of the LLM to guide the exploration of interpretable directed subgraphs in external knowledge graphs, better integrating the two knowledge sources for more accurate reasoning. Extensive experiments on multiple real-world datasets demonstrate the effectiveness of KnowPath. Our code and data are available at https://github.com/tize-72/KnowPath.

Figures (8)

• Figure 1: (a.) The LLMs-only approach suffers from severe hallucinations. (b.) The LLMs with KGs approach provides insufficient information, and their graph-based reasoning with KGs is often inaccurate. (c.) We first mine the internal knowledge of LLMs, offering more information for external KG reasoning and achieving better integration of internal and external knowledge in LLMs.
• Figure 2: The workflow of KnowPath. It contains: (a) Inference Paths Generation to exploit the internal knowledge of LLMs, (b) Subgraph Exploration to generate a interpretable directed subgraph, (c) Evaluation-based Answering to integrate internal and external knowledge.
• Figure 3: Comparison of KnowPath, its individual components, and strong baseline methods (ToG and PoG) on the performance across four commonly used knowledge-based question answering datasets.
• Figure 4: Visualization of the cost-effectiveness analysis on four knowledge-based question-answering datasets.
• Figure 5: Analysis of key parameters.