The recent advancements in the field of 3D Gaussian Splatting (3DGS) have significantly pushed the boundaries of real-time rendering and large-scale scene reconstruction. Innovations such as Hard Gaussian Splatting (HGS) have addressed the issue of spurious artifacts by growing hard Gaussians that fill gaps in classical Gaussian Splatting, leading to superior novel view synthesis (NVS) results. Additionally, frameworks like WRF-GS have revolutionized wireless channel modeling by leveraging 3D Gaussian primitives for efficient wireless radiation field (WRF) reconstruction, outperforming traditional methods in spatial spectrum synthesis and channel state information prediction. Momentum-GS has introduced momentum-based self-distillation to enhance large-scale scene reconstruction, promoting consistency and accuracy across blocks, while decoupling the number of blocks from physical GPU constraints. ResGS has tackled the challenge of detail recovery in 3DGS by introducing a residual densification method, enabling adaptive detail retrieval and progressive refinement. Group Training strategies have optimized training efficiency by organizing Gaussian primitives into manageable groups, achieving faster convergence and improved rendering quality. Lastly, Proc-GS has integrated procedural modeling into the 3DGS framework for scalable and high-fidelity building generation, demonstrating potential for expansive cityscape creation. These developments collectively highlight a shift towards more efficient, scalable, and high-fidelity applications of 3DGS across various domains.