GR-Agent: Adaptive Graph Reasoning Agent under Incomplete Knowledge

2025-12-16

AI Summary

Existing KGQA benchmarks assume complete knowledge graphs (KGs) and evaluate only shallow, single-hop triplet retrieval, overlooking the practical challenge of answering questions over incomplete KGs that require multi-hop reasoning to infer missing facts. Method: We propose the first KGQA framework tailored for incomplete KGs, modeling question answering as an interactive, agent-based reasoning process over graph-structured environments. Our approach employs reinforcement learning to dynamically orchestrate composable graph reasoning operators—including neighbor expansion, path aggregation, and relation inversion—while maintaining a memory of multi-hop relational paths as evidence. Contribution/Results: We introduce a novel benchmark for KGQA on incomplete KGs and design a zero-shot, environment-interaction–driven reasoning agent with memory augmentation. Experiments show our method significantly outperforms non-training baselines on our benchmark, matches supervised models in performance, and maintains robustness across both complete and incomplete KG settings.

Technology Category

• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Natural Language Processing: Question Answering

Application Category

• 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: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Large language models (LLMs) achieve strong results on knowledge graph question answering (KGQA), but most benchmarks assume complete knowledge graphs (KGs) where direct supporting triples exist. This reduces evaluation to shallow retrieval and overlooks the reality of incomplete KGs, where many facts are missing and answers must be inferred from existing facts. We bridge this gap by proposing a methodology for constructing benchmarks under KG incompleteness, which removes direct supporting triples while ensuring that alternative reasoning paths required to infer the answer remain. Experiments on benchmarks constructed using our methodology show that existing methods suffer consistent performance degradation under incompleteness, highlighting their limited reasoning ability. To overcome this limitation, we present the Adaptive Graph Reasoning Agent (GR-Agent). It first constructs an interactive environment from the KG, and then formalizes KGQA as agent environment interaction within this environment. GR-Agent operates over an action space comprising graph reasoning tools and maintains a memory of potential supporting reasoning evidence, including relevant relations and reasoning paths. Extensive experiments demonstrate that GR-Agent outperforms non-training baselines and performs comparably to training-based methods under both complete and incomplete settings.

Problem

• Addresses knowledge graph incompleteness in question answering
• Proposes a benchmark for evaluating reasoning under missing facts
• Introduces an adaptive agent for graph-based inference tasks

Innovation

• Adaptive Graph Reasoning Agent for incomplete knowledge graphs
• Constructs interactive environment from knowledge graph for question answering
• Uses graph reasoning tools and memory for evidence-based inference