Description: This paper reevaluates the performance of Graph Autoencoders for link prediction, showing that with proper hyperparameter tuning, orthogonal embeddings, and linear propagation, a simple GAE can match SOTA GNNs in accuracy while being more efficient.
Description: This paper presents an efficient method for temporal graph link prediction, achieving state-of-the-art results on a large-scale temporal dataset (TGB).
Griffin: Towards a Graph-Centric Relational Database Foundation Model [ICML 2025]
Authors:Yanbo Wang, Xiyuan Wang, Quan Gan, Minjie Wang, Qibin Yang, David Wipf, Muhan Zhang
Description: This paper introduces Griffin, a novel graph-based foundation model that successfully tackles complex relational databases, showing strong predictive performance and transfer learning through its unified data handling, specialized graph neural network, and effective multi-stage pretraining.
An Empirical Study of Realized GNN Expressiveness [ICML 2024]
Authors:Yanbo Wang, Muhan Zhang
Description: This study investigates the capabilities of realized graph neural networks (GNNs), providing insights beyond the general GNN function space.
