Learning Efficient and Generalizable Graph Retriever for Knowledge-Graph Question Answering

2025-06-11

AI Summary

To address the weak generalization and insufficient structural support of graph-based retrievers in Knowledge Graph Question Answering (KGQA), this paper proposes RAPL: (1) a causal-driven two-stage annotation mechanism that explicitly models query-to-subgraph causal relationships; (2) a model-agnostic, universal graph transformation method that jointly encodes intra- and inter-triple interactions; and (3) a path-guided retrieval-reasoning decoupling paradigm, where path encoding enhances structured output. Evaluated on multiple KGQA benchmarks, RAPL outperforms state-of-the-art methods by 2.66%–20.34%, demonstrating significantly improved robustness and generalization across varying model scales and diverse datasets. Moreover, it delivers more interpretable and architecture-compatible structured inputs to downstream reasoning modules.

Technology Category

• Knowledge Representation and Reasoning: Action, Change, and Causality
• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

Large Language Models (LLMs) have shown strong inductive reasoning ability across various domains, but their reliability is hindered by the outdated knowledge and hallucinations. Retrieval-Augmented Generation mitigates these issues by grounding LLMs with external knowledge; however, most existing RAG pipelines rely on unstructured text, limiting interpretability and structured reasoning. Knowledge graphs, which represent facts as relational triples, offer a more structured and compact alternative. Recent studies have explored integrating knowledge graphs with LLMs for knowledge graph question answering (KGQA), with a significant proportion adopting the retrieve-then-reasoning paradigm. In this framework, graph-based retrievers have demonstrated strong empirical performance, yet they still face challenges in generalization ability. In this work, we propose RAPL, a novel framework for efficient and effective graph retrieval in KGQA. RAPL addresses these limitations through three aspects: (1) a two-stage labeling strategy that combines heuristic signals with parametric models to provide causally grounded supervision; (2) a model-agnostic graph transformation approach to capture both intra- and inter-triple interactions, thereby enhancing representational capacity; and (3) a path-based reasoning strategy that facilitates learning from the injected rational knowledge, and supports downstream reasoner through structured inputs. Empirically, RAPL outperforms state-of-the-art methods by $2.66%-20.34%$, and significantly reduces the performance gap between smaller and more powerful LLM-based reasoners, as well as the gap under cross-dataset settings, highlighting its superior retrieval capability and generalizability. Codes are available at: https://github.com/tianyao-aka/RAPL.

Problem

• Enhancing generalization in graph retrievers for KGQA
• Improving structured reasoning with knowledge graphs
• Reducing reliability issues in LLMs via retrieval

Innovation

• Two-stage labeling combines heuristic and parametric models
• Model-agnostic graph transformation captures triple interactions
• Path-based reasoning enhances structured knowledge injection