Research map / Graph learning / GNN architectures
Building powerful and equivariant graph neural networks with structural message-passing
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 equivariant local-context matrices instead of ordinary node features (ablation) | SMP · Graph properties, test set | average log MSE: -2.77 -> -3.59 | Helped |
| Use equivariant local-context matrices instead of ordinary node features (ablation) | Fast SMP · Graph properties, test set | average log MSE: -2.37 -> -3.53 | Helped |
| Use standard SMP rather than Fast SMP | SMP · Graph properties, test set | average log MSE: -3.53 -> -3.59 | No clear effect |
| Use standard SMP rather than Fast SMP | SMP · Geodesic distance, test set | log MSE: -3.31 -> -3.59 | Helped |
| Use Fast SMP rather than standard SMP | Graph property prediction · Graph diameter, test set | log MSE: -3.58 -> -3.65 | Helped |
| Use Fast SMP rather than standard SMP | Graph property prediction · Connectivity, test set | log MSE: -2.97 -> -2.72 | Hurt |
| Use SMP rather than GIN | Cycle detection · Cycle length 8, graph size 72 | test accuracy: 92.5 -> 99.9 | Helped |
| Use SMP rather than GIN | Cycle detection · OOD, cycle length 8, graph size 72 | test accuracy: 88.8 -> 79.5 | Hurt |
| Use SMP rather than GIN with random identifiers | Cycle detection · 6-cycle, graph size 56, 200 training samples | test accuracy: 65.8 -> 87.7 | Helped |
| Use SMP rather than MPNN (ablation) | ZINC regression · ZINC 10k, no edge features | MAE: 0.272 -> 0.219 | Helped |
| Use SMP rather than MPNN (ablation) | ZINC regression · ZINC 10k, with edge features | MAE: 0.189 -> 0.138 | Helped |
Closest papers by meaning
- Schur Nets: exploiting local structure for equivariance in higher order graph neural networks (NeurIPS 2024 · similarity 0.91)
- Are High-Degree Representations Really Unnecessary in Equivariant Graph Neural Networks? (NeurIPS 2024 · similarity 0.90)
- Equivariance Everywhere All At Once: A Recipe for Graph Foundation Models (NeurIPS 2025 · similarity 0.90)
- On the Expressive Power of Sparse Geometric MPNNs (ICLR 2025 · similarity 0.90)
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.89)
- Accurate and Scalable Graph Neural Networks via Message Invariance (ICLR 2025 · 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.