Research map / Graph learning / Graph self-supervised learning
GraphDE: A Generative Framework for Debiased Learning and Out-of-Distribution Detection on Graphs
Zenan Li, Qitian Wu, Fan Nie, Junchi Yan
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 GraphDE-a to TopKPool backbone | TopKPool · SPMotif debiasing | test accuracy (%): 78.43 -> 88.54 | Improved |
| GraphDE-a versus strongest baseline GRAND | TopKPool · SPMotif debiasing | test accuracy (%): 83.25 -> 88.54 | Improved |
| add GRAND to GAT backbone | GAT · SPMotif debiasing | test accuracy (%): 65.02 -> 65.27 | No clear effect |
| add DropEdge to GCN backbone | GCN · SPMotif debiasing | test accuracy (%): 68.67 -> 65.59 | Hurt |
| use EM-based GraphDE-a instead of variational GraphDE-v | GraphDE · SPMotif debiasing, TopKPool | test accuracy (%): 85.11 -> 88.54 | Improved |
| GraphDE detector versus strongest baseline on MNIST-75sp | GLocalKD · MNIST-75sp OOD detection | AUROC (%): 86.22 -> 95.53 | Improved |
| GraphDE versus MSP on MNIST-75sp | MSP · MNIST-75sp OOD detection | AUROC (%): 62.37 -> 95.53 | Improved |
| GraphDE versus GLocalKD on DrugOOD | GLocalKD · DrugOOD OOD detection | FPR95 (%): 70.28 -> 80.30 | Hurt |
| GraphDE versus OCGIN on Collab | OCGIN · Collab OOD detection | FPR95 (%): 86.70 -> 64.40 | Improved |
| remove L_reg (structure estimation model) (ablation) | GraphDE-a · Collab debiasing | test accuracy: over 6% drop at biased ratio 0.25 | Hurt |
| remove L_cls (classification model) (ablation) | GraphDE-a · Collab OOD detection | AUROC: approximately 3% gap at biased ratio 0.3 | Hurt |
| raise prior ratio above 0.7 (ablation) | GraphDE · Collab debiasing | test accuracy | Hurt |
Later papers that compared with this method
1 later accepted paper reports GraphDE as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Breaking One-Size-Fits-All: Revisiting Out-of-Distribution Detection on Graphs Under Diverse Distribution Shifts (AAAI 2026 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GOLD: Graph Out-of-Distribution Detection via Implicit Adversarial Latent Generation (ICLR 2025 · similarity 0.92)
- GOODAT: Towards Test-Time Graph Out-of-Distribution Detection (AAAI 2024 · similarity 0.92)
- Bridging OOD Detection and Generalization: A Graph-Theoretic View (NeurIPS 2024 · similarity 0.90)
- Disentangled Graph Self-supervised Learning for Out-of-Distribution Generalization (ICML 2024 · similarity 0.90)
- Pruning Spurious Subgraphs for Graph Out-of-Distribution Generalization (NeurIPS 2025 · similarity 0.90)
- Graph Out-of-Distribution Detection Goes Neighborhood Shaping (ICML 2024 · 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.
