The field of reinforcement learning and robotic control is rapidly advancing, with a focus on improving sample efficiency, robustness, and generalizability. Recent developments have highlighted the importance of model-based approaches, which leverage environment models to optimize policy performance. Additionally, there is a growing interest in offline reinforcement learning, which aims to learn from historical data without requiring access to the environment. Researchers are also exploring the use of skills and hierarchies to improve policy transfer and adaptability in complex, long-horizon tasks. Noteworthy papers in this area include the introduction of the Conditional Diffusion Model Planner, which enables accurate modeling of environmental dynamics and planning of superior policies. The NeoRL-2 benchmark has also been proposed, which provides a more realistic and challenging evaluation platform for offline reinforcement learning algorithms. Furthermore, the Quality-focused Active Adversarial Policy has been developed, which improves safety in human-robot interaction by actively mitigating the risk of grasping the human hand. Other notable contributions include the development of discrete diffusion skills for offline reinforcement learning, model-based offline reinforcement learning with adversarial data augmentation, and state-aware perturbation optimization for robust deep reinforcement learning.