The field of novel view synthesis is witnessing significant advancements with the introduction of innovative methods based on Gaussian Splatting. Researchers are exploring ways to improve the quality of synthesized views, particularly in sparse-view scenarios where traditional methods struggle. A key direction in this area is the development of techniques that incorporate epipolar depth priors, geometric priors, and constrained diffusion priors to enhance the accuracy and robustness of Gaussian Splatting-based approaches. These advancements have led to state-of-the-art performance in various benchmarks and datasets. Noteworthy papers in this area include NexusGS, which introduces a novel point cloud densification strategy, and SparseGS-W, which enables the reconstruction of complex outdoor scenes with as few as five training images. Additionally, CoMapGS and EVPGS propose innovative approaches to address region-specific uncertainty and extrapolated view synthesis, respectively.