Research map / Graph learning / GNN architectures
Labeling Trick: A Theory of Using Graph Neural Networks for Multi-Node Representation Learning
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 DRNL labeling trick to GCN (GCN+DRNL vs GCN) | GCN · ogbl-ppa test | Hits@100 (%): 18.67 -> 45.24 | Helped |
| add SEAL (DRNL + SortPooling) vs GCN | GCN · ogbl-citation2 test | MRR (%): 84.74 -> 87.67 | Helped |
| add labeling trick (SEAL vs GCN) | GCN · ogbl-collab test | Hits@50 (%): 47.14 -> 64.74 | Helped |
| add labeling trick (SEAL vs GraphSAGE/GCN+LRGA) | plain GAE · ogbl-ddi test | Hits@20 (%): 62.30 -> 30.56 | Hurt |
| add DRNL labeling to GCN | GCN · ogbl-citation2 test | MRR (%): 84.74 -> 81.27 | Hurt |
| use zero-one labeling instead of all-zero (no labeling) (ablation) | GCN (1-hop) · ogbl-collab test | Hits@50 (%): 25.92 -> 44.79 | Helped |
| use zero-one labeling instead of all-zero (no labeling) (ablation) | GCN (1-hop) · ogbl-citation2 test | MRR (%): 36.98 -> 38.78 | Helped |
| use DRNL instead of zero-one labeling (ablation) | GCN (1-hop) · ogbl-citation2 test | MRR (%): 38.78 -> 81.27 | Helped |
| use GIN instead of GCN with zero-one labeling (ablation) | zero-one labeling trick · ogbl-collab test | Hits@50 (%): 44.79 -> 59.48 | Helped |
| use DRNL instead of DE labeling | GCN · ogbl-ppa test | Hits@100 (%): 36.48 -> 45.24 | Helped |
Closest papers by meaning
- Distance Encoding: Design Provably More Powerful Neural Networks for Graph Representation Learning (NeurIPS 2020 · similarity 0.91)
- Node Identifiers: Compact, Discrete Representations for Efficient Graph Learning (ICLR 2025 · similarity 0.91)
- NLGT: Neighborhood-based and Label-enhanced Graph Transformer Framework for Node Classification (AAAI 2025 · similarity 0.90)
- H$^3$GNNs: Harmonizing Heterophily and Homophily in GNNs via Self-Supervised Node Encoding (ICLR 2026 · similarity 0.90)
- Combining Label Propagation and Simple Models out-performs Graph Neural Networks (ICLR 2021 · similarity 0.90)
- Why Propagate Alone? Parallel Use of Labels and Features on Graphs (ICLR 2022 · 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.