The field of 3D Gaussian Splatting (3DGS) is moving towards addressing long-standing issues such as false transparency, geometric accuracy, and robustness under various conditions. Researchers are exploring innovative solutions, including noise-guided splatting, hierarchical training approaches, and structured representations, to improve the quality and consistency of 3D reconstructions. Notable papers in this area include Fix False Transparency by Noise Guided Splatting, which proposes a method to reduce false transparency in 3DGS, and GSPlane, which introduces a structured representation for planar regions to improve geometric accuracy. Additionally, the development of new benchmarks, such as Raindrop GS, and filtering methods, like Extreme Views, are pushing the boundaries of 3DGS under challenging conditions.