The field of 3D scene reconstruction is moving towards addressing complex challenges such as extreme motion blur and dynamic scene reconstruction. Researchers are exploring innovative approaches to decouple dynamic and static components within scenes, and to integrate geometric and generative priors for improved realism and consistency. Notable papers include GeMS, which proposes a framework for 3D Gaussian Splatting that reconstructs scenes directly from extremely blurred images, and ExtraGS, which introduces a holistic framework for trajectory extrapolation that integrates geometric and generative priors. DriveSplat is also noteworthy for its high-quality reconstruction method for driving scenarios based on neural Gaussian representations with dynamic-static decoupling.