Research map / Graph learning / GNN architectures
GLAD: Bidirectional Structure-Attribute Alignment via Latent Graph Diffusion Models
Jiankai Zuo, Yu Zhang, Yang Zhang, Zihao Yao, Yaying Zhang
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 backward topological alignment loss (lambda > 0) (ablation) | GLAD · Cora, Citeseer, Amazon-C, Amazon-P attribute recovery | Recall@10: removal drops Recall@10 by nearly 30% | Improved |
| add structure-aware classifier-free guidance at sampling (ablation) | GLAD · Citeseer attribute recovery | Recall@k | Improved |
| diffuse in decoupled latent spaces instead of a joint concatenated space (ablation) | GLAD · four graph datasets | Recall@k | Improved |
| replace adversarial alignment (SAT) with latent diffusion for sample diversity | attribute completion · Cora, 80% missing | average pairwise distance: 0.18 -> 1.12 (ground truth 1.18) | Improved |
| add backward alignment so generated attributes reconstruct topology | attribute completion · Cora, 80% missing | link prediction AUC (%): 81.5% -> 88.4% vs WGNN | Improved |
| use GLAD instead of SAT-GAT for node classification on Amazon-Photo | completed attributes · Amazon-Photo | accuracy (%): 92.50 -> 92.41 | No clear effect |
| use latent diffusion completion instead of GAN-based SAT for traffic ETA | traffic sensor networks · London ETA | MAE: 97.74 -> 80.01 | Improved |
| use GLAD for congestion classification vs best baseline | traffic networks · Madrid, Melbourne, London | accuracy (%): Madrid 90.23 -> 93.38 | Improved |
Closest papers by meaning
- GLAD: Improving Latent Graph Generative Modeling with Simple Quantization (AAAI 2025 · similarity 0.87)
- On the Benefits of Attribute-Driven Graph Domain Adaptation (ICLR 2025 · similarity 0.87)
- Dissecting the Failure of Invariant Learning on Graphs (NeurIPS 2024 · similarity 0.87)
- When Edge Independence Fails: Joint Graph Diffusion with Latent Sociability Priors (NeurIPS 2026 · similarity 0.86)
- What Is Missing For Graph Homophily? Disentangling Graph Homophily For Graph Neural Networks (NeurIPS 2024 · similarity 0.86)
- Graph Structure Refinement with Energy-based Contrastive Learning (AAAI 2025 · similarity 0.86)
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.
