Research map / Graph learning / GNN architectures
Ordered GNN: Ordering Message Passing to Deal with Heterophily and Over-smoothing
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 ordered message-passing gates instead of independently predicted gates (ablation) | Ordered GNN · CiteSeer, homophily, 8 layers | accuracy: 75.78{\scriptstyle\pm1.49} -> 76.45{\scriptstyle\pm1.35} | Helped |
| use ordered message-passing gates instead of independently predicted gates (ablation) | Ordered GNN · Wisconsin, heterophily, 8 layers | accuracy: 86.27{\scriptstyle\pm5.23} -> 85.69{\scriptstyle\pm4.90} | Hurt |
| use ordered message-passing gates instead of independently predicted gates (ablation) | Ordered GNN · CiteSeer, over-smoothing, 32 layers | accuracy: 79.30 -> 79.40 | Helped |
| add independently predicted gates to the mean-aggregator GNN (ablation) | Bare GNN · CiteSeer, homophily, 8 layers | accuracy: 73.79{\scriptstyle\pm1.57} -> 75.78{\scriptstyle\pm1.49} | Helped |
| add independently predicted gates to the mean-aggregator GNN (ablation) | Bare GNN · Wisconsin, heterophily, 8 layers | accuracy: 53.92{\scriptstyle\pm5.79} -> 86.27{\scriptstyle\pm5.23} | Helped |
| add independently predicted gates to the mean-aggregator GNN (ablation) | Bare GNN · CiteSeer, over-smoothing, 32 layers | accuracy: 74.70 -> 79.30 | Helped |
| add differentiable SOFTOR to ordered gates (ablation) | Ordered GNN · CiteSeer, homophily, 8 layers | accuracy: 76.45{\scriptstyle\pm1.35} -> 77.31{\scriptstyle\pm1.73} | Helped |
| add differentiable SOFTOR to ordered gates (ablation) | Ordered GNN · Wisconsin, heterophily, 8 layers | accuracy: 85.69{\scriptstyle\pm4.90} -> 88.04{\scriptstyle\pm3.63} | Helped |
| add differentiable SOFTOR to ordered gates (ablation) | Ordered GNN · CiteSeer, over-smoothing, 32 layers | accuracy: 79.40 -> 80.80 | Helped |
| use Ordered GNN rather than the strongest reported baseline | node classification · CiteSeer | accuracy: 78.02{\scriptstyle\pm1.15} -> 77.31\pm1.73 | Hurt |
| use Ordered GNN rather than the strongest reported baseline | node classification · ogbn-arxiv | accuracy: 73.11{\scriptstyle\pm0.21} -> 72.81{\scriptstyle\pm0.21} | Hurt |
| use Ordered GNN rather than the strongest reported baseline | node classification · PubMed, 64 layers | accuracy: 91.60 -> 91.60 | No clear effect |
Closest papers by meaning
- GAMMA: Gated Multi-hop Message Passing for Homophily-Agnostic Node Representation in GNNs (NeurIPS 2025 · similarity 0.90)
- Sign is Not a Remedy: Multiset-to-Multiset Message Passing for Learning on Heterophilic Graphs (ICML 2024 · similarity 0.89)
- Mitigating Oversmoothing Through Reverse Process of GNNs for Heterophilic Graphs (ICML 2024 · similarity 0.88)
- Towards Dynamic Message Passing on Graphs (NeurIPS 2024 · similarity 0.88)
- Accurate and Scalable Graph Neural Networks via Message Invariance (ICLR 2025 · similarity 0.88)
- Making Classic GNNs Strong Baselines Across Varying Homophily: A Smoothness–Generalization Perspective (NeurIPS 2025 · similarity 0.88)
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.