Minwen Liao

I am a research intern at the University of Washington in Prof. Dieter Fox’s group and a senior at Xinjiang University. Previously, I worked on robotic manipulation at Tsinghua AIR.

I am seeking PhD opportunities for Fall 2027.

Minwen Liao

My research focuses on how robots can learn from visual experience and use it to guide their actions. I am particularly interested in:

World-action models and representations

How can we learn representations that capture physical dynamics and task-relevant information, supporting both prediction and action?

In-context robot learning

How can robots adapt to new tasks using human and robot demonstrations as context?

From understanding to control

How can we translate the visual understanding of foundation models into reliable robot actions, using feedback to adapt and recover from mistakes?

I’m always happy to discuss research and potential collaborations—feel free to reach out!

News

  • [2025.11] One paper accepted to WACV 2026 !
  • [2025.11] One paper accepted to AAAI 2026 !
  • [2025.08] One paper accepted to CoRL 2025 !
  • [2025.07] Our team HURRICANE won the First Prize in RoboCon 2025 !
  • [2025.06] One paper accepted to ICCV 2025 !

Research

Representative papers are highlighted.

Flex-π: A Multi-Stream World-Action Model with Compute Flexibility
Ge Yan*, Jinghao Liu*, Yuzhi Fan*, Lei Cai, Minwen Liao, Jesse Zhang†, Dieter Fox†
Preprint, 2026

* Equal contribution    † Equal advising


We introduce Flex-π, a multi-stream world-action model that jointly predicts RGB, 3D geometry, object semantics, and actions while supporting flexible inference modes from a single checkpoint.

GM-MoE: Low-Light Enhancement with Gated-Mechanism Mixture-of-Experts
Minwen Liao*, Hao Bo Dong*, Xinyi Wang*, Kurban Ubul†, Yihua Shao†, Ziyang Yan†
ICCV, 2025

* Equal contribution    † Corresponding authors


We propose GM-MoE, a novel approach for low-light image enhancement using gated-mechanism mixture-of-experts architecture.

RoboChemist: Long-Horizon and Safety-Compliant Robotic Chemical Experimentation
Zongzheng Zhang*, Chenghao Yue*, Haobo Xu*, Minwen Liao, Xianglin Qi, Huan-ang Gao, Ziwei Wang†, Hao Zhao†
CoRL, 2025

* Equal contribution    † Corresponding authors


We propose RoboChemist, a framework for autonomous robotic chemical experimentation with long-horizon planning and safety compliance.

Beyond Research

My journey into robotics began with the RoboCon competition, where I developed my passion for computer vision and embodied intelligence.

Beyond research, I enjoy playing the violin 🎻, running marathons 🏃, and capturing moments through photography 📷. These hobbies inspire my creativity and keep me balanced.

A glimpse into my life beyond research

Photography 1 Photography 2
Photography 3 Photography 4
Photography 5 Photography 6