The field of humanoid robotics and manipulation is rapidly evolving, with a focus on developing more efficient, robust, and adaptable control policies. Recent research has emphasized the importance of integrating learning-based methods with model-based approaches to reduce training complexity and ensure safety and stability. Notable developments include the use of adversarial policy learning, physics-informed world models, and sparse anchor posture curriculum learning to improve whole-body control, non-prehensile manipulation, and motion synthesis. These innovations have the potential to significantly advance the field and enable more sophisticated humanoid robot capabilities. Noteworthy papers include: Adversarial Locomotion and Motion Imitation for Humanoid Policy Learning, which proposes a novel framework for adversarial policy learning between upper and lower body. PIN-WM: Learning Physics-INformed World Models for Non-Prehensile Manipulation, which presents a physics-informed world model for robust policy learning and generalization. ProMoGen: Progressive Motion Generation via Sparse Anchor Postures Curriculum Learning, which introduces a novel framework for integrating trajectory guidance with sparse anchor motion control.