Research map / Graph learning / GNN architectures
LogicXGNN: Grounded Logical Rules for Explaining Graph Neural Networks
Chuqin Geng, Ziyu Zhao, Zhaoyue Wang, Haolin Ye, Yuhe Jiang, Xujie Si
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| evaluate fidelity on final grounded explanations instead of concept space | GraphTrail · Mutagenicity | fidelity (%): 73.90 -> 65.93 | Hurt |
| replace GraphTrail with LogicXGNN rules and grounding | GraphTrail · BBBP | Fid_D (%): 50.00 -> 85.32 | Improved |
| replace GLGExplainer with LogicXGNN | GLGExplainer · IMDB-BINARY | Fid_D (%): 0.00 -> 74.17 | Improved |
| use LogicXGNN instead of GraphTrail | GraphTrail · NCI1 | Fid_D (%): 60.04 -> 73.81 | Improved |
| use LogicXGNN instead of GraphTrail | GraphTrail · Mutagenicity | validity (%): 61.90 -> 100.00 | Improved |
| use LogicXGNN instead of GraphTrail | GraphTrail · BBBP | coverage class 0 (%): 0.00 -> 47.86 | Improved |
| use LogicXGNN instead of GraphTrail | GraphTrail · Mutagenicity | stability (%): 37.50 -> 66.67 | Improved |
| increase decision tree rule depth from 5 to 10 (ablation) | LogicXGNN · BAMultiShapes | rule recovery | Improved |
| use LogicXGNN instead of baselines at scale | GLGExplainer, GraphTrail · Reddit, Twitch, Github | scalability | Improved |
| use decision-tree grounding rules instead of single representative subgraph per concept | LogicXGNN vs baselines · 10 to 100x runtime | runtime: 10-100x faster | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GraphTrail — GraphTrail: Translating GNN Predictions into Human-Interpretable Logical Rules (NeurIPS 2024) · better on 13 of 13 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Towards Robust Fidelity for Evaluating Explainability of Graph Neural Networks (ICLR 2024 · similarity 0.91)
- Reconsidering Faithfulness in Regular, Self-Explainable and Domain Invariant GNNs (ICLR 2025 · similarity 0.91)
- Beyond Topological Self-Explainable GNNs: A Formal Explainability Perspective (ICML 2025 · similarity 0.90)
- GraphTrail: Translating GNN Predictions into Human-Interpretable Logical Rules (NeurIPS 2024 · similarity 0.90)
- Stratified GNN Explanations through Sufficient Expansion (AAAI 2024 · similarity 0.89)
- EiG-Search: Generating Edge-Induced Subgraphs for GNN Explanation in Linear Time (ICML 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
