Advances in differentiable modelling and artificial intelligence are creating new opportunities to improve how hydrologic processes are represented, diagnosed, and learned from data. This session welcomes contributions spanning process-based, data-driven, physics-informed, and hybrid hydrologic modelling. Topics include differentiable model structures, large-sample learning and regionalisation, sensitivity analysis, uncertainty quantification, diagnosis of model misspecification, and methods for discovering or refining process representations. We particularly encourage studies that use artificial intelligence to identify missing or inadequate processes, develop interpretable model components, quantify predictive and structural uncertainty, or improve transfer across locations, time periods, and hydrologic regimes. Contributions on numerical methods, analytic and automatic differentiation, trustworthy evaluation, and applications across spatial and temporal scales are also welcome. By bringing together hydrologists, hydroinformaticians, and machine-learning researchers, the session aims to advance models that are computationally efficient, scientifically interpretable, and better able to characterise uncertainty and structural limitations.