At the 2026 World Artificial Intelligence Conference (WAIC), WeRide released the physical AI cognitive foundation large model WIIT, exploring a new direction for artificial intelligence to evolve from data understanding to real-world cognition.

According to the introduction, WIIT proposes the concept of "Minimum Physical Fact Unit (Physical Fact)" based on real-world scenarios. It breaks down continuously changing real environments into identifiable and verifiable basic fact units, and constructs an AI understanding framework for the physical world based on this.

Based on physical facts, WIIT has developed four core capabilities: fact extraction, fact reasoning, fact verification, and fact orchestration. Among them, fact extraction is used to identify key elements from complex environments; fact reasoning helps the model understand the relationships between different facts; fact verification enhances the AI's ability to judge the accuracy of real-world information; and fact orchestration supports the model in organizing tasks and generating decisions based on facts.

WeRide stated that the core challenge of physical AI lies in enabling models not only to process information but also to understand spatial relationships, behavioral logic, and dynamic changes in the real world. By building a cognitive system based on physical facts, WIIT aims to enhance AI's environmental understanding and decision-making capabilities in embodied intelligence scenarios such as autonomous driving and robotics.