Research map / Graph learning / Graph self-supervised learning
Reconstruction Outcomes Look Similar but Processes Differ: Improving Context Consistency and Coverage in Graph Masked Auto-Encoder
Geng Tang, Keyu Liu, Xibei Yang, Yuhua Qian
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 density-partitioned masking instead of uniform masking (ablation) | C2-GMAE · Chameleon | accuracy (%): 80.59 -> 82.42 | Improved |
| add density-partitioned masking instead of uniform masking (ablation) | C2-GMAE · Squirrel | accuracy (%): 74.68 -> 74.80 | No clear effect |
| add heterophilic edges amplification (ablation) | C2-GMAE · Chameleon | accuracy (%): 80.90 -> 82.42 | Improved |
| add heterophilic edges amplification (ablation) | C2-GMAE · BlogCatalog | accuracy (%): 89.22 -> 90.38 | Improved |
| add PE amplification on heterophilic edges (ablation) | C2-GMAE · BlogCatalog | accuracy (%): 87.85 -> 90.38 | Improved |
| add attention amplification (heterophily gate) (ablation) | C2-GMAE · Actor | accuracy (%): 39.72 -> 42.59 | Improved |
| add noise amplification on masked nodes (ablation) | C2-GMAE · BlogCatalog | accuracy (%): 89.78 -> 90.38 | Improved |
| add loss amplification on heterophilic edges (ablation) | C2-GMAE · Squirrel | accuracy (%): 74.40 -> 74.80 | Improved |
| increase attention amplification gamma beyond 0.1 (ablation) | C2-GMAE · Actor | accuracy (%) | Hurt |
| set PE amplification beta larger than 0.1 (ablation) | C2-GMAE · Actor | accuracy (%) | Hurt |
| use C2-GMAE vs best baseline on homophilic WikiCS | C2-GMAE · WikiCS | accuracy (%): 79.38 -> 79.59 | No clear effect |
| use C2-GMAE vs GraphPAE on heterophilic Actor | C2-GMAE · Actor | accuracy (%): 38.55 -> 42.59 | Improved |
Closest papers by meaning
- Teacher-guided Edge Discriminator for Personalized Graph Masked Autoencoder (AAAI 2025 · similarity 0.89)
- Rethinking Graph Masked Autoencoders through Alignment and Uniformity (AAAI 2024 · similarity 0.89)
- Self-supervised Masked Graph Autoencoder via Structure-aware Curriculum (ICML 2025 · similarity 0.89)
- GraphCroc: Cross-Correlation Autoencoder for Graph Structural Reconstruction (NeurIPS 2024 · similarity 0.88)
- Graph Positional and Structural Encoder (ICML 2024 · similarity 0.85)
- Linear Transformer Topological Masking with Graph Random Features (ICLR 2025 · similarity 0.85)
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.
