Unitree Robotics founder Wang Xingxing said at the 2026 World Robot Conference that the company is advancing self-evolving physical AI robot models by using large AI models to auto-develop control code, which will be verified through evaluation and manual review. The system will incorporate simulation, real-world and human behavior data for continuous improvement as robot deployment scales. The firm also demonstrated robots evolving from single to group intelligence through multi-machine coordination, showing cluster tasks performed by humanoid, quadrupedal and wheel-legged models along with the new GD01 manned exoskeleton. Wang Xingxing noted the biggest hurdle for embodied AI is insufficient generalization ability (around 80% task completion via voice commands in unfamiliar settings) which could be achieved in 2 to 3 years or take 5 to 10 years otherwise, adding that current models struggle with correcting tactile errors in fine operations due to input output discrepancies.