On April 10, Pony.ai unveiled its latest breakthrough in the field of physical AI—the PonyWorld World Model 2.0. Compared to version 1.0, the 2.0 model features self-diagnosis and targeted evolution capabilities, enabling it to autonomously identify weaknesses, evolve in a targeted manner, and guide human teams in research and development as well as data collection. The model achieves three core breakthroughs in capabilities: automated retrospective attribution analysis for each driving decision; automatic identification of scenarios with insufficient accuracy and generation of targeted data collection tasks; and automatic generation of tailored training scenarios based on the model's weak points in vehicle-end performance. Currently, the World Model 2.0 has been integrated into Pony.ai's L4 autonomous driving fleet and R&D system, enhancing vehicle safety, comfort, and traffic efficiency.