Bridging graph structure and knowledge-guided editing for interpretable temporal knowledge graph reasoning

Shiqi Fan

doi:10.1016/j.neunet.2026.108811

Bridging graph structure and knowledge-guided editing for interpretable temporal knowledge graph reasoning

2026-03-05

0

PRE

AI

DOI：10.1016/j.neunet.2026.108811

Original

Original request for help

Save

Abstract

En 中文

• We propose IGETR, the first path-refine reasoning framework integrating temporal graph neural networks with knowledge-augmented LLMs, effectively combining structural reasoning and semantic refinement to improve logical consistency and interpretability. • We design a three-stage pipeline that grounds reasoning in graph structures, refines paths via LLM editing, and aggregates multi-hop evidence with a graph Transformer, ensuring data-driven reliability while enabling knowledge-guided, controllable reasoning. • We validate the superiority of IGETR through experiments on three TKG datasets, demonstrating its effectiveness in addressing the key challenges of TKGR, with ablation studies validate the effectiveness of key components in the framework.

Keywords:

IGETR

temporal knowledge graph reasoning

graph neural networks

knowledge-augmented LLMs

interpretable reasoning

Bridging graph structure and knowledge-guided editing for interpretable temporal knowledge graph reasoning

Abstract

Journal

Organization