Xinyao Li

Robotics Institute

Carnegie Mellon University

xinyaol2@andrew.cmu.edu

xinyaoli511@gmail.com

I am an incoming MSR student at the Robotics Institute, Carnegie Mellon University, where I will be advised by Prof. Guanya Shi.

I received my BS in Computer Science from Shanghai Jiao Tong University, in the ACM Honors Class — a selective program admitting 30 students per year.

My research is on humanoid robots — in particular how to make whole-body policies do something useful in contact with the physical world, rather than merely reproducing human-like poses. Recent work spans contact-controllable object interaction (ContactMimic), unified loco-manipulation control (ULTRA), and real-time human-humanoid interaction (RHINO). I train in simulation (Isaac Lab, MuJoCo, IsaacGym) and deploy on real hardware, mostly the Unitree G1.

Before CMU I was a research intern at the RoboVision @ Illinois Lab, University of Illinois Urbana-Champaign, advised by Prof. Saurabh Gupta, where I did ContactMimic. Earlier I worked on efficient vision-language-action models at ByteDance Research (Robotics Group), and on humanoid interaction at the APEX Lab, SJTU, advised by Prof. Weinan Zhang.

news

Aug 25, 2026	Starting at the CMU Robotics Institute as an MSR student, advised by Prof. Guanya Shi. 🎒
Aug 25, 2026	RHINO and ULTRA are both accepted to IROS 2026. 🎉
Jul 11, 2026	ContactMimic is on arXiv — humanoid object interaction with explicit part-level contact control. Project page
Jan 22, 2026	Manipulation as in Simulation (Camera Depth Models) is accepted to ICLR 2026.
Jun 01, 2025	Started as a research intern at the RoboVision @ Illinois Lab, UIUC, advised by Prof. Saurabh Gupta.

selected publications

ContactMimic: Humanoid Object Interaction via Contact Control

Xinyao Li, Xialin He, Runpei Dong, and Saurabh Gupta

arXiv 2026

Xinyao Li and Xialin He contributed equally. University of Illinois Urbana-Champaign.

Abs arXiv Bib HTML Code

Keypoint tracking alone is insufficient for object interaction tasks such as sitting on a chair, wiping a board, or pushing furniture, where the robot can reach the correct pose without making meaningful physical contact with the object. We present ContactMimic, a learning framework that tracks explicit part-level binary contact commands alongside keypoint trajectories. ContactMimic is made possible through the use of contact-following rewards and a trajectory augmentation scheme aimed at breaking the correlations between keypoint trajectories and contact labels. The resulting policy successfully decouples contact behavior from keypoint geometry, and achieves precise physical contact as well as contact-controllability (produce or suppress contact during deployment as desired). Simulation experiments across 10 diverse human-object interaction motions confirm that ContactMimic exhibits contact controllability that enables it to complete manipulation tasks without task-specific rewards, while also outperforming keypoint-only trackers on contact-relevant tasks. Ablations confirm the necessity of the proposed trajectory augmentation scheme and sim2real deployment validates contact controllability in the real world across 5 different motions.
@article{li2026contactmimic, title = {ContactMimic: Humanoid Object Interaction via Contact Control}, author = {Li, Xinyao and He, Xialin and Dong, Runpei and Gupta, Saurabh}, journal = {arXiv 2026}, year = {2026}, month = jul, note = {Xinyao Li and Xialin He contributed equally. University of Illinois Urbana-Champaign.} }
RHINO: Learning Real-Time Humanoid-Human-Object Interaction from Human Demonstrations

Jingxiao Chen, Xinyao Li, Jiahang Cao, Zhengbang Zhu, Wentao Dong, Minghuan Liu, Ying Wen, Yong Yu, Liqing Zhang, and Weinan Zhang

IROS 2026

Jingxiao Chen, Xinyao Li and Jiahang Cao contributed equally.

Abs arXiv Bib HTML Video Code

Propose the first real-time humanoid interaction framework capable of learning from human demonstrations, enabling dynamic task-switching and immediate responses to human instructions.
@inproceedings{chen2026rhino, title = {RHINO: Learning Real-Time Humanoid-Human-Object Interaction from Human Demonstrations}, author = {Chen, Jingxiao and Li, Xinyao and Cao, Jiahang and Zhu, Zhengbang and Dong, Wentao and Liu, Minghuan and Wen, Ying and Yu, Yong and Zhang, Liqing and Zhang, Weinan}, booktitle = {IROS 2026}, year = {2025}, month = feb, note = {Jingxiao Chen, Xinyao Li and Jiahang Cao contributed equally.} }