Yichu Yang
Papers
1
Total Citations
7
H-Index
1
About
Yichu Yang is a leading researcher at the forefront of embodied AI and robot manipulation, with a focus on building generalist agents that can operate in the real world. Yang’s most notable contribution is the development of **GR-2**, a pioneering generative video-language-action model that leverages web-scale knowledge for robotic control. This work represents a paradigm shift in robotics: by pre-training on an unprecedented dataset of **38 million Internet video clips** and over **50 billion tokens**, GR-2 learns the fundamental dynamics of the physical world before being fine-tuned for specific manipulation tasks. This approach has yielded a state-of-the-art generalist robot agent capable of versatile and generalizable manipulation, earning **7 citations** in its first year. Yang’s research bridges the gap between large-scale vision-language models and practical robotics, demonstrating how web-scale video pre-training can imbue robots with a rich understanding of object interactions and task sequences. By moving beyond narrow, task-specific models, Yang is helping to define a new generation of robots that can adapt to novel environments and objects, making significant strides toward truly autonomous and intelligent robotic assistants.
Research Focus
Key Achievements
Top Papers
- 1