[Apple Research Team Discloses Progress in Robot "Brain" Development: Average Success Rate of 91% in Single-Task Tests] Golden Finance reported on November 7th that Apple's research team disclosed progress in its robot "brain" research. On November 4th local time, the Apple research team submitted a paper titled "Perceptive Humanoid Motion from Simulation" to the preprint platform arXiv, disclosing the HALO autonomous control system for humanoid robots. The research team also released code for data generation, model training, and deployment, covering motion generation, motion tracking, task orchestration, and simulation and real robot operation. According to the paper, the research team conducted 360 single-task tests in six real-world environments that were not pre-scanned or mapped, achieving an average success rate of 91%. In another real-world environment, they conducted 10 tests on three combined tasks, achieving an average success rate of 70%. The success rate for the combination "bending over an obstacle—walking towards a target—pushing a box" was 50%. It is reported that the research team used 250 NVIDIA H100 GPUs and spent approximately 48 hours synthesizing the main training data. The motion generator was trained through imitation learning, using 64 B200 frames; the motion tracker was trained through reinforcement learning and policy distillation, using 4 H100 frames. It should be noted that the simulated data generation did not completely eliminate human intervention. The team stated that new physical tasks still required manually written prompts and designed motion keyframes, and some actions also utilized human video or motion capture footage. The research team also pointed out that currently, robots mainly rely on visual perception and have difficulty directly perceiving contact; when objects leave the camera's field of view or are occluded, performance in tasks involving more contact deteriorates. Incorporating tactile perception is one direction for future improvements.