The field of robot manipulation is moving towards more precise and autonomous control, with a focus on developing frameworks and methods that can learn from demonstrations and adapt to new situations. Researchers are exploring the use of reinforcement learning, imitation learning, and hybrid approaches to improve the efficiency and accuracy of robot manipulation. The development of large-scale datasets and taxonomies for robot manipulation is also a key area of research, enabling more flexible and autonomous laboratory automation. Notable papers in this area include: RM-RL, which introduces a role-model reinforcement learning framework that unifies online and offline training in real-world environments, achieving significant gains in real-world manipulation. DexCanvas, which presents a large-scale hybrid real-synthetic human manipulation dataset that can facilitate research in robotic manipulation learning and contact-rich control. TARMAC, which introduces a taxonomy for robot manipulation in chemistry, providing a structured foundation for more flexible and autonomous laboratory automation. NeuralTouch, which integrates neural descriptor fields and tactile sensing to enable accurate and generalizable grasping through gentle physical interaction. FieldGen, which enables scalable, diverse, and high-quality real-world data collection with minimal human supervision, achieving higher success rates and improved stability compared to teleoperation-based baselines.