The field of 3D point cloud understanding and tracking is rapidly advancing, with a focus on developing more efficient and accurate methods for processing and analyzing large-scale 3D data. Recent research has explored the use of novel frameworks and architectures, such as mixture-of-experts models and state space models, to improve the performance of 3D understanding and tracking tasks. Additionally, there is a growing interest in developing methods that can adapt to changing environments and distributions, such as test-time adaptation and online learning approaches. These advances have the potential to enable more accurate and efficient 3D perception and understanding in a variety of applications, including robotics, autonomous driving, and augmented reality. Notable papers in this area include: DoReMi, which proposes a domain-representation mixture framework for generalizable 3D understanding, and MambaTrack3D, which introduces a state space model framework for LiDAR-based object tracking under high temporal variation. CompTrack is also noteworthy, as it presents a novel end-to-end framework for 3D single object tracking that systematically eliminates both spatial and informational redundancy in point clouds.