The field of graph neural networks and temporal reasoning is rapidly advancing, with a focus on improving the efficiency and effectiveness of models in capturing complex interactions and temporal dependencies. Researchers are exploring innovative architectures and techniques, such as sparse graph convolutional networks, disentangled multi-span evolutionary networks, and higher-order structure temporal graph neural networks, to enhance performance in tasks like human action recognition, temporal knowledge graph reasoning, and link prediction. Notable papers in this area include: A paper proposing a novel sparse ST-GCNs generator, which achieves comparable performance to dense models with significantly fewer parameters. A paper introducing a disentangled multi-span evolutionary network for temporal knowledge graph reasoning, demonstrating substantial performance improvements over state-of-the-art methods. A paper proposing a higher-order structure temporal graph neural network, which effectively captures group interactions and reduces memory costs during training.