Research map / Graph learning / GNN architectures
A Complete Expressiveness Hierarchy for Subgraph GNNs via Subgraph Weisfeiler-Lehman Tests
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 |
|---|---|---|---|
| add single-point aggregation agg^P_vv to GNN-SSWL (ablation) | GNN-SSWL · Counting Substructure: Triangle | MAE: 0.0098 -> 0.0064 | Helped |
| add single-point aggregation agg^P_vv to GNN-SSWL (ablation) | GNN-SSWL · Counting Substructure: Tailed Tri. | MAE: 0.0090 -> 0.0067 | Helped |
| add single-point aggregation agg^P_vv to GNN-SSWL (ablation) | GNN-SSWL · Counting Substructure: Star | MAE: 0.0089 -> 0.0078 | Helped |
| add single-point aggregation agg^P_vv to GNN-SSWL (ablation) | GNN-SSWL · Counting Substructure: 4-Cycle | MAE: 0.0107 -> 0.0079 | Helped |
| add single-point aggregation agg^P_vv to GNN-SSWL (ablation) | GNN-SSWL · Counting Substructure: 5-Cycle | MAE: 0.0142 -> 0.0108 | Helped |
| add single-point aggregation agg^P_vv to GNN-SSWL (ablation) | GNN-SSWL · Counting Substructure: 6-Cycle | MAE: 0.0189 -> 0.0154 | Helped |
| add single-point aggregation agg^P_vv to GNN-SSWL (ablation) | GNN-SSWL · ZINC Subset | MAE: 0.082±0.003 -> 0.070±0.005 | Helped |
| add single-point aggregation agg^P_vv to GNN-SSWL (ablation) | GNN-SSWL · ZINC Full | MAE: 0.026±0.001 -> 0.022±0.002 | Helped |
| use GNN-SSWL+ rather than the strongest listed prior node-based subgraph GNN | Counting Substructure: 6-Cycle · Counting Substructure: 6-Cycle | MAE: 0.0550 -> 0.0154 | Helped |
| use GNN-SSWL+ rather than the strongest listed prior subgraph GNN | ZINC Full · ZINC Full | MAE: 0.024±0.003 -> 0.022±0.002 | Helped |
Closest papers by meaning
- Beyond Weisfeiler-Lehman: A Quantitative Framework for GNN Expressiveness (ICLR 2024 · similarity 0.94)
- An Empirical Study of Realized GNN Expressiveness (ICML 2024 · similarity 0.92)
- Balancing Efficiency and Expressiveness: Subgraph GNNs with Walk-Based Centrality (ICML 2025 · similarity 0.91)
- Towards Bridging Generalization and Expressivity of Graph Neural Networks (ICLR 2025 · similarity 0.91)
- On the Rademacher Complexity of Graph Neural Networks: Unifying Expressivity and Geometry (NeurIPS 2026 · similarity 0.91)
- Weisfeiler-Leman at the margin: When more expressivity matters (ICML 2024 · similarity 0.91)
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.