The field of humanoid robotics is moving towards developing more robust and safe control systems. Researchers are focusing on creating frameworks that enable humanoid robots to predict and respond to potential falls, minimizing damage to the robot's components. Another area of focus is on achieving precise motion with strong force interaction, which is crucial for tasks that require intense external force interaction. The development of dual-agent reinforcement learning control frameworks and kinematics-aware multi-policy reinforcement learning is allowing for more stable and proactive force interaction in high-load industrial scenarios. Notable papers include: SafeFall, which demonstrates significant performance improvements in reducing peak contact forces and joint torques, and HAFO, which achieves stable control of humanoid robots under various strong force interactions. Kinematics-Aware Multi-Policy Reinforcement Learning for Force-Capable Humanoid Loco-Manipulation is also noteworthy for its decoupled three-stage training pipeline, enabling the robot to actively exert and regulate interaction forces with the environment.