WeRide Introduces WITT, a Physical AI Cognitive Foundation Model Buil…

WeRide unveils WITT, a Physical AI foundation model that extracts and verifies atomic physical facts from autonomous driving data, reducing token costs by up to 98% compared to general-purpose models. The system processes real-world operational data into trusted learning signals through fact extraction, reasoning, verification and curation. If you're building autonomous driving or robotics systems, the cost structure of your training pipeline just shifted dramatically: fact-based models beat general-purpose ones by 98% on token costs. Your data becomes your competitive moat only if you can extract and verify it at scale.