Research map / Graph learning / GNN architectures
Evaluating Attribution for Graph Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use CAM instead of gradient-based attribution methods | GCN/MPNN/Graph Nets attribution · Graph attribution tasks (classification and regression) | attribution accuracy (AUROC / Kendall tau): p<0.05 for 90% of pairwise comparisons | Helped |
| use GCN instead of MPNN or Graph Nets as the model | attribution methods · Graph attribution tasks | mean attribution accuracy: p<0.05 for 31.7% of pairwise comparisons | Helped |
| use Graph Nets architecture instead of simpler GCN | attribution methods · Graph attribution tasks | attribution accuracy variability: p<0.05 for 20% of pairwise comparisons | Hurt |
| use CAM for attribution under label noising | GCN/MPNN/GAT · Noised-label faithfulness | PARC: p<0.05 for 94% of comparisons | Helped |
| use CAM or IG under spurious-correlation handicapping | GCN/MPNN · Spurious correlation P&Q faithfulness | PARC: p<0.05 for 48.8% of comparisons | Helped |
| use CAM instead of other methods under graph perturbations | GNN attribution · Perturbed test graphs | variance of change in attribution AUROC: p<0.05 for 71.7% of comparisons | Helped |
| use attention weights as attribution for GAT instead of gradient methods | GAT · Graph attribution tasks | attribution accuracy | Hurt |
| train longer (later epochs) | GNN attribution · Graph attribution tasks | attribution accuracy | Hurt |
Closest papers by meaning
- GOAt: Explaining Graph Neural Networks via Graph Output Attribution (ICLR 2024 · similarity 0.90)
- quanda: An Interpretability Toolkit for Training Data Attribution Evaluation (NeurIPS 2026 · similarity 0.88)
- Faithful and Accurate Self-Attention Attribution for Message Passing Neural Networks via the Computation Tree Viewpoint (AAAI 2025 · similarity 0.87)
- Benchmarking the Attribution Quality of Vision Models (NeurIPS 2024 · similarity 0.87)
- $\texttt{dattri}$: A Library for Efficient Data Attribution (NeurIPS 2024 · similarity 0.87)
- Towards Robust Fidelity for Evaluating Explainability of Graph Neural Networks (ICLR 2024 · similarity 0.87)
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.