The field of dynamic graph learning and temporal modeling is witnessing significant advancements, driven by the development of innovative architectures and techniques. Researchers are exploring new ways to effectively model temporal relationships, capture spatial-temporal correlations, and improve the accuracy and efficiency of predictive models. Notably, there is a growing interest in leveraging graph neural networks, attention mechanisms, and transfer learning to address complex tasks such as traffic prediction, link prediction, and influence maximization. These advancements have the potential to positively impact various applications, including transportation systems, financial risk control, and social network analysis. Noteworthy papers include: STEI-PCN, which proposes an efficient pure convolutional network for traffic prediction via spatial-temporal encoding and inferring, achieving competitive computational efficiency and superior performance on most evaluation metrics. Between Linear and Sinusoidal, which introduces a simpler alternative to sinusoidal time encoders, demonstrating improved performance and significant savings in model parameters. Detecting Credit Card Fraud via Heterogeneous Graph Neural Networks, which proposes a method based on heterogeneous graph neural networks with graph attention, outperforming existing GNN models on the IEEE-CIS Fraud Detection dataset. GC-GAT, which presents a lane graph-based motion prediction model that predicts graph-based goal proposals and fuses them with cross attention, achieving state-of-the-art results on the nuScenes motion prediction dataset. Trajectory Encoding Temporal Graph Networks, which introduces automatically expandable node identifiers as learnable temporal positional features, effectively balancing transductive accuracy with inductive generalisation.