Tag
This paper introduces a logic-based framework for extracting global logical rules from Simple Graph Convolution networks using minimal abductive explanations, aiming to provide compact and faithful explanations for node classification in graph neural networks.
Introduces a benchmark to evaluate how knowledge editing methods handle logical consequences of fact edits, revealing that existing approaches like ROME and FT accurately insert direct assertions but fail to propagate entailed knowledge, with a performance gap of up to 24%.
This paper introduces GRiD, a framework that uses diffusion models and reinforcement learning to generate graph-like rules (e.g., cycles, branches) for knowledge graph reasoning, addressing the limitations of existing chain-rule mining methods. Experiments on six benchmarks show competitive performance in KG completion tasks.