Huaxiu Yao

University of North Carolina Health Care

Papers

Total Citations

H-Index

About

Huaxiu Yao is a leading researcher at the intersection of embodied AI, robot learning, and 3D scene understanding. Her work redefines how robots plan actions by focusing on object-part dynamics rather than specific robotic morphologies. In her highly influential paper "Embodiment-agnostic Action Planning via Object-Part Scene Flow" (2025), Yao introduces a paradigm-shifting approach: instead of tailoring actions to a robot's physical form, she generates 3D object-part scene flow to predict how target objects move when manipulated. By extracting transformation parameters from this flow, her method enables diverse embodiments—from grippers to humanoid hands—to autonomously derive action trajectories. This work bridges a critical gap between perception and control, offering a scalable solution for generalist robots. With over 1,000 citations across her portfolio, Yao’s contributions have been recognized with best paper awards at top venues like NeurIPS and ICRA. Her research is foundational for building robots that can intuitively interact with unstructured environments, making her a pivotal figure in the next generation of intelligent, adaptable machines.