ResNets of All Shapes and Sizes: Convergence of Training Dynamics in the Large-scale Limit

Collection of recent arXiv (20 Mar 2026) machine-learning and statistical-learning papers covering: rigorous convergence of ResNet training dynamics in the joint infinite-depth/width/embedding limit (MLU regime) with tight error rates; pitfalls and a principled framework for data preparation in imbalanced insurance settings (support points, Chatterjee correlation, missing-data handling) integrated into an InsurAutoML pipeline; a hybrid conditional diffusion + DeepONet surrogate (cDDPM-DeepONet) for high-fidelity stress prediction in hyperelastic materials; multi-domain causal empirical Bayes (