Research map / Graph learning / GNN architectures
Combining Label Propagation and Simple Models out-performs 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 |
|---|---|---|---|
| Add Correct and Smooth to an MLP base predictor | MLP · Products | accuracy: 63.41 -> 84.18 | Helped |
| Add Correct and Smooth with FDiff-scale to a linear base predictor | Linear · Products | accuracy: 50.05 -> 83.01 | Helped |
| Use training and validation labels in C&S instead of training labels alone | C&S · ogbn-arxiv | accuracy: 72.62 -> 74.02 | Helped |
| Remove final prediction smoothing from C&S (ablation) | C&S FDiff-scale + MLP · Products | accuracy: 72.72 -> 84.18 | Hurt |
| Remove both label-based propagation steps (ablation) | MLP base predictor · Products | accuracy: 80.34 -> 84.18 | Hurt |
| Replace Autoscale with FDiff-scale (ablation) | C&S + MLP · Products | accuracy: 63.41 -> 84.18 | Helped |
| Use FDiff-scale rather than Autoscale (ablation) | C&S + Linear · Rice31 | accuracy: 86.42 -> 87.55 | Mixed |
| Use FDiff-scale rather than Autoscale (ablation) | C&S + MLP · US County | accuracy: 89.64 -> 89.85 | Helped |
| Use the simple LP method instead of a learned GNN | Transductive node classification · Cora | accuracy: 85.77 -> 86.50 | Helped |
| Apply C&S to existing GNN predictors | GCNII or GAT · ogbn-arxiv | accuracy | Helped |
| Use a linear base predictor with C&S instead of the Products SOTA GNN | Products · Products | training time: over 100 times faster | Helped |
Closest papers by meaning
- Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence (ICML 2025 · similarity 0.91)
- Classic GNNs are Strong Baselines: Reassessing GNNs for Node Classification (NeurIPS 2024 · similarity 0.91)
- Label-free Node Classification on Graphs with Large Language Models (LLMs) (ICLR 2024 · similarity 0.90)
- NoisyGL: A Comprehensive Benchmark for Graph Neural Networks under Label Noise (NeurIPS 2024 · similarity 0.90)
- NLGT: Neighborhood-based and Label-enhanced Graph Transformer Framework for Node Classification (AAAI 2025 · similarity 0.90)
- Diffusing to the Top: Boost Graph Neural Networks with Minimal Hyperparameter Tuning (ICLR 2025 · similarity 0.90)
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.