LogicXGNN: Grounded Logical Rules for Explaining Graph Neural Networks
Published in ICLR, 2026
Recommended citation: Geng, C., Zhao, Z., Wang, Z., Ye, H., Jiang, Y., & Si, X. LogicXGNN: Grounded Logical Rules for Explaining Graph Neural Networks. In International Conference on Learning Representations (ICLR), 2026. https://openreview.net/pdf?id=nqZe43tRY9
LogicXGNN extracts grounded logical rules to explain Graph Neural Networks, using graph-hashing-based predicate construction from local subgraph structure and GNN activation patterns.