The field of structured data learning is witnessing significant growth, with a focus on extending the benefits of large-scale pretraining to tabular domains. Recent developments have centered around creating unified libraries and frameworks that standardize workflows, provide consistent access to state-of-the-art models, and enable efficient fine-tuning and evaluation. These advancements aim to address the challenges of heterogeneous preprocessing pipelines, fragmented APIs, and inconsistent fine-tuning procedures. Furthermore, innovations in tabular in-context learning have led to the development of architectures that capture hierarchical feature interactions, scalable attention mechanisms, and bidirectional information flow. In the realm of visual object representation learning, unified libraries have emerged, providing a scalable foundation for rapid experimentation and efficient transfer of research advances to real-world applications. Noteworthy papers include: Orion-MSP, which introduces a multi-scale sparse attention mechanism for tabular in-context learning, and TabTune, which presents a unified library for inference and fine-tuning tabular foundation models. DORAEMON is also notable, as it unifies visual object modeling and representation learning across diverse scales.