The field of autonomous systems and 3D scene understanding is rapidly evolving, with a focus on developing more accurate and efficient methods for perception, prediction, and control. Researchers are exploring innovative approaches to integrate probabilistic models, Bayesian estimation, and language guidance to improve the safety and adaptability of autonomous systems. Additionally, there is a growing interest in developing vision-only methods for 3D scene understanding, which can overcome the limitations of traditional LiDAR-based approaches. These advancements have the potential to enable more robust and flexible autonomous systems, with applications in areas such as driving, robotics, and surveillance. Notable papers in this area include: ShelfOcc, which introduces a native 3D supervision method for vision-based occupancy estimation, and CylinderDepth, which proposes a novel geometry-guided method for multi-view consistent self-supervised surround depth estimation. POMA-3D is also a noteworthy paper, as it presents a self-supervised 3D representation model learned from point maps, which can serve as a strong backbone for various 3D understanding tasks.