The field of simulations and video generation is moving towards more realistic and accurate representations of complex environments and tasks. Researchers are focusing on developing physics-based models that can handle large-scale simulations and interactions with deformable objects. Another notable trend is the use of geometric inductive biases and algebraic frameworks to improve the efficiency and accuracy of robot learning and video generation tasks. Noteworthy papers in this area include the introduction of Chrono::CRM, a general-purpose simulation solution for terramechanics problems, and MedGen, a medical video generation model that achieves leading performance in visual quality and medical accuracy. The hybrid diffusion policy approach hPGA-DP also shows promise in improving task performance and training efficiency, while Geometry Forcing provides a simple yet effective method for encouraging video diffusion models to internalize latent 3D representations.