The recent developments in the field of graph-based machine learning have shown a significant shift towards addressing complex and practical challenges. There is a notable emphasis on advancing unsupervised and federated learning techniques, particularly in scenarios where data privacy and distribution are critical concerns. Innovations in contrastive learning have been leveraged to enhance the robustness and accuracy of graph-level representations, especially in federated settings where data cannot be centrally shared. Additionally, there is a growing focus on anomaly detection in multiplex graphs, where methods are being developed to handle multiple interaction types and improve unsupervised anomaly scoring. Scalability remains a key area of improvement, with new approaches being introduced to efficiently construct embeddings for large attributed graphs, supporting a variety of downstream tasks. These advancements collectively push the boundaries of what is achievable in graph-based learning, particularly in real-world, data-sensitive applications.