The field of 3D generative models is rapidly advancing, with a focus on developing more sophisticated and geometry-aware techniques. Recent research has centered around improving the quality and diversity of generated 3D objects, as well as enabling more effective manipulation and editing of these objects. One key direction is the development of part-level generation methods, which allow for the creation of 3D objects with multiple, distinct parts that can be edited independently. Another important area of research is the incorporation of geometric constraints and priors into generative models, enabling the production of 3D objects that are more realistic and consistent with human expectations. Additionally, there is a growing interest in developing methods that can align generated 3D objects with specific orientations and poses, as well as techniques for generating 3D objects that are consistent with user-specified geometric instructions. Overall, these advances have the potential to enable a wide range of applications, from computer-aided design and robotics to video games and virtual reality. Notable papers in this area include PartCrafter, which introduces a compositional latent diffusion transformer for part-aware 3D generation, and DreamCS, which proposes a geometry-aware text-to-3D generation framework with unpaired 3D reward supervision. Harmonizing Geometry and Uncertainty: Diffusion with Hyperspheres is also worth mentioning, as it presents a novel approach to preserving class geometry in diffusion models.