Xiaomi-Robotics-1 shows that more data beats bigger models when training robots to move

Xiaomi-Robotics-1 is a new robotics training method proving that dataset scale beats model size. Researchers trained the model using over 100,000 hours of human motion data captured via handheld grippers instead of physical robots. This scaling approach yields continuous performance improvements without hitting immediate diminishing returns, offering a new blueprint for robotic locomotion training.

Opening Kapyn…