The field of 3D scene reconstruction and denoising is experiencing significant advancements, driven by the development of innovative deep learning frameworks and techniques. Researchers are focusing on improving the accuracy and efficiency of scene reconstruction, particularly in scenarios with limited camera motion or high levels of noise. Novel approaches, such as the integration of diffusion priors and reward-guided Gaussian integration, are being explored to address the challenges of expansive reconstruction. Additionally, graph-based methods and geometric attention mechanisms are being utilized to enhance the robustness and consistency of depth denoising algorithms. These advancements have the potential to significantly impact various applications, including sensor simulators, 3D object recognition, and autonomous systems. Noteworthy papers include: RGE-GS, which achieves state-of-the-art performance in reconstruction quality through a novel expansive reconstruction framework, and DMD-Net, which presents a deep learning framework for mesh denoising that obtains competitive or better results compared to existing methods. Consistent Time-of-Flight Depth Denoising via Graph-Informed Geometric Attention is also notable for its ability to enhance temporal stability and spatial sharpness in ToF depth denoising.