Yutong Liang

San Diego, CA

Robot Learning · Character Animation

Research

Hi there! I'm Yutong Liang, a graduate student in Computer Science and Engineering at the University of California San Diego, advised by Prof. Xiaolong Wang.

I build dexterous robot learning systems both in simulation and reality. My current work combines human demonstrations, world models, and reinforcement learning for dynamic, contact-rich manipulation.

About Me

Selected Publications

arXiv·2026

FACT: Failure-Aware Causal Training for World-Action Models

Quanquan Peng*, Yutong Liang*, Rui Yan, Nicklas Hansen, Xiaolong Wang

Takeaway: Failure rollouts teach an action-conditioned world model to predict the consequences of failed actions without imitating them.

WEBSITE PAPER CODE VIDEO

@article{peng2026fact,
      title={FACT: Failure-Aware Causal Training for World-Action Models},
      author={Quanquan Peng and Yutong Liang and Rui Yan and Nicklas Hansen and Xiaolong Wang},
      journal={arXiv preprint arXiv:2608.10232},
      year={2026}
}

ECCV·2026

ConTrack: Constrained Hand Motion Tracking with Adaptive Trade-off Control

Yutong Liang, Quanquan Peng, Ri-Zhao Qiu, Xiaolong Wang

Takeaway: When transferring human hand-object manipulation to a robot, the primary goal is to preserve the object's motion, then the remaining control flexibility can be used to replicate human hand-object contact.

WEBSITE PAPER CODE VIDEO X POST

@article{liang2026contrack,
      title={ConTrack: Constrained Hand Motion Tracking with Adaptive Trade-off Control}, 
      author={Yutong Liang and Quanquan Peng and Ri-Zhao Qiu and Xiaolong Wang},
      journal={arXiv preprint arXiv:2606.03177},
      year={2026}
}

CVPR·2026·Highlight

XL-VLA: Cross-Hand Latent Representation for Vision-Language-Action Models

Guangqi Jiang*, Yutong Liang*, Jianglong Ye, Jia-Yang Huang, Changwei Jing, Rocky Duan, Pieter Abbeel, Xiaolong Wang†, Xueyan Zou†

Takeaway: Embodiment-invariant latent action space enhances performance as demonstrations scale across different hand embodiments, similarly to scaling with additional data from a single hand.

WEBSITE PAPER CODE X POST

@article{jiang2026crosshand,
      title={Cross-Hand Latent Representation for Vision-Language-Action Models}, 
      author={Guangqi Jiang and Yutong Liang and Jianglong Ye and Jia-Yang Huang and Changwei Jing and Rocky Duan and Pieter Abbeel and Xiaolong Wang and Xueyan Zou},
      journal={arXiv preprint arXiv:2603.10158},
      year={2026}
}

Eurographics·2026

DexterCap: An Affordable and Automated System for Capturing Dexterous Hand-Object Manipulation

Yutong Liang*, Shiyi Xu*, Yulong Zhang*, Bowen Zhan, He Zhang, Libin Liu

Takeaway: Dexterous in-hand manipulation can be captured by providing dense motion information while minimizing interference caused by markers.

WEBSITE PAPER CODE VISUALIZATION DATA X POST

@article{liang2026dextercap,
      title={DexterCap: An Affordable and Automated System for Capturing Dexterous Hand-Object Manipulation}, 
      author={Yutong Liang and Shiyi Xu and Yulong Zhang and Bowen Zhan and He Zhang and Libin Liu},
      journal={arXiv preprint arXiv:2601.05844},
      year={2026}
}

All Publications

Selected Writing

All Writing

Education

University of California San Diego

M.S. in Computer Science and Engineering

Sep. 2025 - Present

San Diego, United States

Peking University

B.S. in Computer Science and Technology

Sep. 2021 - Jul. 2025

Beijing, China