The field of computational fluid dynamics and related areas is witnessing a significant shift towards physically consistent modeling and simulation. Researchers are increasingly focusing on developing methods that not only provide accurate predictions but also adhere to the underlying physical laws, ensuring reliability and robustness in various applications. This trend is evident in the development of novel frameworks and techniques that synergize machine learning with traditional solver-based approaches, enabling stable and scalable simulations. Furthermore, there is a growing emphasis on uncertainty quantification and mitigation, leading to more accurate and trustworthy forecasts. The integration of diffusion models and preference optimization is also being explored to improve the performance of precipitation nowcasting and other applications. Noteworthy papers in this area include: Physically consistent neural operators that enforce physical constraints and quantify uncertainties, achieving high-fidelity spatiotemporal predictions. SynCast, a method that employs diffusion sequential preference optimization to align conflicting metrics and achieve superior performance in precipitation nowcasting. Error bounded compression methods, such as EBCC, that provide heavily compressed weather and climate datasets while maintaining physical consistency and accuracy. Hybrid reconstruction frameworks that combine the efficiency of linear reconstruction with the robustness of nonlinear formulations, reducing computational cost without compromising accuracy or stability.