On November 7, Apple's research team published a paper on the preprint platform arXiv titled "Perceptive Humanoid Motion from Simulation," unveiling HALO, an autonomous control system for humanoid robots. The team deployed the system on the Unitree G1 humanoid robot for validation, enabling motions learned in a simulated environment to be transferred to unfamiliar real-world rooms for execution. They also released the code for data generation, model training, and deployment, covering motion generation, tracking, task orchestration, as well as simulation and real-machine operation. Tests showed that in six real environments that had not been pre-scanned or mapped, the robot completed 360 individual tasks with an average success rate of 91%. In another real environment, three combined tasks were each tested 10 times, yielding an average success rate of 70%; among them, the combined task of "bending down to pass an obstacle—walking toward the target—pushing the box" achieved a success rate of 50%.