The field of robotic manipulation and perception is rapidly advancing, with a focus on developing innovative solutions for complex tasks such as task-oriented grasping, geometric assembly, and articulated object reconstruction. Researchers are exploring new approaches to improve the accuracy and efficiency of robotic systems, including the use of zero-shot and one-shot learning, explicit modeling of physical interactions, and generative priors for hand-object interactions. Notable trends include the development of unified frameworks for modeling multi-solid systems and articulated objects, as well as the application of domain adaptation algorithms for semantic segmentation in industrial environments. Noteworthy papers include Unisoma, which presents a unified Transformer-based model for handling variable numbers of solids, and BiAssemble, which introduces a real-world benchmark for geometric assembly with long-horizon action sequences. UAD is also notable for its unsupervised affordance distillation method, which enables task-conditioned affordance models to generalize to unseen object instances and task instructions.