Generating Graph-like Rules for Knowledge Graph Reasoning via Diffusion Models

2026-05-28

AI Summary

Existing methods for knowledge graph rule mining are constrained by chain-like structures, limiting their ability to model cyclic and branched relational patterns, and further hindered by combinatorial explosion and non-differentiable rule evaluation metrics. This work introduces diffusion mechanisms into graph-structured logical rule generation for the first time, proposing a two-stage training framework: it first employs supervised pre-training to learn prior meta-graph structures, then integrates reinforcement learning with policy gradients to directly optimize the quality of non-differentiable rules. By transcending the limitations of chain-based rules, the approach achieves competitive knowledge graph completion performance across six benchmark datasets. Ablation studies confirm that graph-structured rules significantly enhance both reasoning capability and robustness.

Abstract

Logical rules constitute a cornerstone of knowledge graph (KG) reasoning, valued for their interpretability and ability to model relational patterns. However, existing rule mining methods predominantly focus on simple chain-like rules and therefore neglect the richer relational information encoded in graph-like structures, such as cycles and branches. This limitation is further exacerbated by computational bottlenecks caused by the combinatorial explosion of the search space, which is especially challenging for graph-like rules. Meanwhile, generative approaches such as diffusion models, despite their success in other domains, can not be directly applied to rule mining because their training objectives are not aligned with the goal of learning high-quality rules, and non-differentiable KG rule quality metrics cannot directly guide model optimization. To address these limitations, we propose GRiD, a framework that reformulates graph-like rule discovery as a discrete generative process conditioned on the target relation. GRiD employs a two-phase training strategy. First, supervised pre-training enables GRiD to capture structural priors from subgraphs sampled from the KG meta-graph. Subsequently, reinforcement learning is applied to fine-tune GRiD through policy gradient optimization guided directly by non-differentiable rule-quality metrics. Experiments on six benchmark datasets show that GRiD achieves competitive performance on KG completion tasks. Ablation studies confirm the efficiency and robustness of GRiD and further show that graph-like rules complement chain-like rules in KG completion. Our codes and datasets are available in https://github.com/Haoxiang-Cheng/GRiD

Problem

• knowledge graph reasoning
• graph-like rules
• rule mining
• diffusion models
• combinatorial explosion

Innovation

• graph-like rules
• diffusion models
• knowledge graph reasoning
• reinforcement learning
• rule mining