The recent advancements in the field of 3D reconstruction and rendering have shown a significant shift towards leveraging Gaussian Splatting (3DGS) for high-fidelity, real-time applications. Researchers are focusing on integrating 3DGS into various frameworks to enhance efficiency, accuracy, and robustness across different scenarios, including multi-agent collaboration, dynamic scenes, and large-scale free camera trajectories. Key innovations include the development of adaptive and motion-aware Gaussian representations, which allow for more precise and efficient rendering, even in the presence of irregular camera trajectories or dynamic elements. Additionally, the use of event cameras and multi-temporal granularity fusion is emerging as a powerful tool for handling motion blur and improving deblurring performance in real-world applications. These advancements not only push the boundaries of what is achievable in real-time 3D reconstruction but also pave the way for more integrated and scalable solutions in AR/VR, robotics, and medical imaging.

Noteworthy contributions include MAC-Ego3D, which sets new standards for real-time multi-agent collaboration in 3D reconstruction, and RP-SLAM, which demonstrates state-of-the-art performance in photorealistic SLAM using 3D Gaussian Splatting. SplineGS and SweepEvGS also stand out for their innovative approaches to dynamic scene reconstruction and event-based rendering, respectively.