Foundational physical world model speeds real-world robot interactions
InternW0: A Foundational Physical World Model for Efficient Real-World Interactions
RoboticsArtificial Intelligence
Summary
Understanding how the physical world changes is important for robots to act effectively in it, especially when conditions keep changing. The authors created InternW0, a new model that combines video and robot control data to predict future movements and help robots make decisions faster and more accurately. This model works even when the robot has limited information and faces forces or touch-based interactions. They trained the model on thousands of hours of robot and first-person video data and tested it on both simulated tasks and real-world scientific workflows, such as chemical synthesis and precision pipetting.
What this means in practice
- •For robotics engineers: Enable robots to perform complex multi-step scientific experiments involving contact and force sensing with improved speed and accuracy.
- •For automation developers: Create adaptive control systems for robots that handle real-world unpredictable environments efficiently using integrated video and sensory inputs.
- •For industrial process teams: Implement robot-assisted workflows in chemical manufacturing with precise, contact-aware operations guided by rich physical world modeling.$Commercial implications: This model supports commercial robotic automation products for chemical synthesis, offering more reliable and efficient manufacturing steps.
Authors
Jisong Cai, Yao Mu, Ganlin Yang, Zhe Cao, Zhangzheng Tu, Xing Gao, Kailin Li, Xinyu Zhan, Lixin Yang, Yangkun Zhu, Haoxiang Ma, Ming Zhou, Qiaojun Yu, Yufei Xue, Liqun He, Yifei Yao, Yifan Zhu, Long Ling, Bingqi Jiang, Haoyu Guo, Xueyue Zhu, Bowen Zhou, Bin Zhao, Tianfan Xue, Chunhua Shen, Weinan Zhang
Abstract
Physical intelligence requires more than predicting how the world may evolve: predictions must remain actionable as the world continues to change. We introduce InternW0, the first instantiation of the InternW physical world model series from Shanghai AI Laboratory, built around omnimodal interfaces, asynchronous multi-frequency processing, and local physical modeling under partial observations and external influences. InternW0 jointly learns future visual dynamics and continuous robot control through an asymmetric video--action architecture with flow matching. A high-capacity video expert provides longer-horizon predictive context, while a lightweight action expert operates at a faster timescale. Instead of regenerating the future for every action update, InternW0 reuses layerwise K/V and adapts it to newly observed states through observation-conditioned context routing. Domain-specific interfaces and soft prompts support heterogeneous embodiments, while contact-aware post-training incorporates force and tactile signals for contact-rich manipulation. We train InternW0 on approximately 7,200 hours of heterogeneous robot and egocentric data, including EgoLab, a 275-hour real-laboratory egocentric dataset. Evaluation spans simulation benchmarks and real-world scientific tasks, including a 15-stage metal--organic framework synthesis workflow and 5-stage contact- and force-aware dexterous manipulation for general-purpose quantitative pipetting. These results advance scalable, asynchronous, and science-native physical world models for universal and efficient real-world interactions.