Research map / Graph learning / GNN architectures
Multi-Scale Distillation from Multiple 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 |
|---|---|---|---|
| use multi-scale GAT teachers with attention allocation rather than the strongest single-teacher distillation baseline | MSKD with GAT · PPI | F1-score: 96.83 -> 97.31 | Helped |
| use multi-scale GAT teachers with attention allocation rather than the strongest single-teacher distillation baseline | MSKD with GAT · CiteSeer | F1-score: 94.65 -> 95.47 | Helped |
| use multi-scale GAT teachers with attention allocation rather than the strongest single-teacher distillation baseline | MSKD with GAT · PubMed | F1-score: 89.92 -> 91.81 | Helped |
| use multi-scale GCN teachers with attention allocation rather than the strongest distillation baseline | MSKD with GCN · PPI | F1-score: 61.79 -> 63.67 | Helped |
| use multi-scale GCN teachers with attention allocation rather than the strongest distillation baseline | MSKD with GCN · CiteSeer | F1-score: 88.87 -> 89.67 | Helped |
| use multi-scale GCN teachers with attention allocation rather than the strongest distillation baseline | MSKD with GCN · PubMed | F1-score: 84.13 -> 84.89 | Helped |
| replace equal teacher weights with attention-based weights (ablation) | MSKD with GAT · PPI | F1-score: 97.24 -> 97.31 | Helped |
| replace equal teacher weights with attention-based weights (ablation) | MSKD with GAT · CiteSeer | F1-score: 95.30 -> 95.47 | Helped |
| replace equal teacher weights with attention-based weights (ablation) | MSKD with GAT · PubMed | F1-score: 91.67 -> 91.81 | Helped |
| add a 1-layer GAT teacher to a 2-layer GAT teacher (ablation) | MSKD · PPI | F1-score: 95.96 -> 96.96 | Helped |
| add a 3-layer GAT teacher to a 2-layer GAT teacher (ablation) | MSKD · PPI | F1-score: 95.96 -> 96.92 | Helped |
| add a 3-layer GAT teacher to a 1-layer GAT teacher (ablation) | MSKD · PPI | F1-score: 95.55 -> 96.35 | Helped |
Closest papers by meaning
- Large Language Model Meets Graph Neural Network in Knowledge Distillation (AAAI 2025 · similarity 0.90)
- VQGraph: Rethinking Graph Representation Space for Bridging GNNs and MLPs (ICLR 2024 · similarity 0.89)
- Preference-driven Knowledge Distillation for Few-shot Node Classification (NeurIPS 2025 · similarity 0.89)
- TINED: GNNs-to-MLPs by Teacher Injection and Dirichlet Energy Distillation (ICML 2025 · similarity 0.88)
- Graph Distillation with Eigenbasis Matching (ICML 2024 · similarity 0.88)
- Neural Collapse Inspired Knowledge Distillation (AAAI 2025 · similarity 0.87)
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.