Papers
8
Total Citations
223
H-Index
4
About
Haoshu Fang is a leading researcher in robotic manipulation, with a focus on dexterous grasping, imitation learning, and generalizable policy learning. Their work bridges perception and action, notably through the development of the EyeSight Hand—a fully-actuated dexterous hand with integrated vision-based tactile sensors—and AnyGrasp, a robust grasp perception system operating across spatial and temporal domains. Fang has made foundational contributions to large-scale robotic learning, co-authoring the landmark Open X-Embodiment paper (119 citations), which introduced the RT-X models and datasets that have become a cornerstone for generalist robot policies. They also led the RH20T dataset (55 citations), enabling one-shot imitation learning for diverse skills. Their recent work on CAGE introduces causal attention mechanisms for data-efficient, generalizable manipulation, while RISE demonstrates how 3D perception simplifies real-world imitation learning. With over 200 total citations and multiple publications in top robotics venues, Fang’s research is shaping the future of robots that can learn, adapt, and manipulate objects with human-like dexterity and generalization.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3RISE: 3D Perception Makes Real-World Robot Imitation Simple and Effective22 citations · 2024
- 4
- 5
- 6
- 7A Surprisingly Efficient Representation for Multi-Finger Grasping2 citations · 2024
- 8