The field of graph neural networks (GNNs) is moving towards improving their performance and robustness on real-world graph data. Researchers are exploring innovative methods to leverage classical algorithms and vision models to enhance the capabilities of GNNs. The integration of classical algorithmic priors into GNNs has shown promising results, while vision models have demonstrated remarkable potential for graph structural understanding, particularly in tasks requiring global topological awareness and scale-invariant reasoning. Furthermore, advancements in graph learning, pruning, and quantization are enabling the application of GNNs to larger and more complex graphs. Noteworthy papers include: Leveraging Classical Algorithms for Graph Neural Networks, which demonstrates the effectiveness of pretraining GNNs on classical algorithms for molecular property prediction tasks. The Underappreciated Power of Vision Models for Graph Structural Understanding, which introduces a new benchmark, GraphAbstract, to evaluate models' ability to perceive global graph properties and shows that vision models outperform GNNs on tasks requiring holistic structural understanding.