Tianyou Wang ็Ž‹ๅคฉไฝ‘

I am a first-year PhD student at the Applied AI Lab, Oxford Robotics Institute, University of Oxford, supervised by Prof. Ingmar Posner. My research interests include robot learning and world model for robotic manipulation, with a focus on enabling robots to learn from human data and adapt to diverse, unstructured environments.

Previously, I earned my master's degree in Robotics at the GRASP Laboratory at the University of Pennsylvania, where I worked with Prof. Dinesh Jayaraman and Prof. Kostas Daniilidis. I received my bachelor's degree in Intelligent Manufacturing from SHIEN-MING WU School of Intelligent Engineering, South China University of Technology.

I am always happy to chat about research and explore potential collaborations! Feel free to reach out at tianyou@robots.ox.ac.uk!


News
2026
Had a great time exhibiting with the Oxford Robotics Institute at the Royal Society Summer Science Exhibition 2026 in London โ€” demoing our robots to the public and chatting with visitors of all ages about robot learning! ๐Ÿค–
Jul
My first project during PhD, GLAM, has been out on arXiv! ๐ŸŒ
Jun
I am attending ICRA 2026 in Vienna. Happy to connect!
Jun
I am attending 7th UK Robot Manipulation Workshop in Edingurgh. Happy to connect!
Jan
2025
Our work Reisom has been out on arXiv! ๐Ÿ–
Dec
I started my PhD at the University of Oxford!
Oct
Zeromimic won the Best Paper Award ๐Ÿ† at CVPR 2025 Workshop on 3D Vision Language Models for Robotic Manipulation: Opportunities and Challenges!
Jun
I graduated from Penn! ๐ŸŽˆ๐Ÿ‘จโ€๐ŸŽ“
May
Our work Zeromimic has been accepted to ICRA 2025! ๐ŸŽ‰ My first project in robot learning, and the one that sparked my interest in learning from human videos.
Jan
Selected Publications (view all )
Imitation from Heterogeneous Demonstrations using Grounded Latent-Action World Models
Imitation from Heterogeneous Demonstrations using Grounded Latent-Action World Models

Tianyou Wang, Anson Lei, Joe Watson, Ingmar Posner

Preprint 2026

We introduce GLAM (Grounded Latentโ€‘Action World Models), a pair of latent-action generative models that leverage heterogeneous demonstrations for supervising downstream imitation learning.

Imitation from Heterogeneous Demonstrations using Grounded Latent-Action World Models

Tianyou Wang, Anson Lei, Joe Watson, Ingmar Posner

Preprint 2026

We introduce GLAM (Grounded Latentโ€‘Action World Models), a pair of latent-action generative models that leverage heterogeneous demonstrations for supervising downstream imitation learning.

Reisom: Zero-shot Reconstruction of In-Scene Object Manipulation from Video
Reisom: Zero-shot Reconstruction of In-Scene Object Manipulation from Video

Dixuan Lin*, Tianyou Wang*, Zhuoyang Pan, Yufu Wang, Lingjie Liu, Kostas Daniilidis (* equal contribution)

Preprint 2025

We introduce Reisom, a system for reconstructing in-scene object manipulation from a single monocular RGB video.

Reisom: Zero-shot Reconstruction of In-Scene Object Manipulation from Video

Dixuan Lin*, Tianyou Wang*, Zhuoyang Pan, Yufu Wang, Lingjie Liu, Kostas Daniilidis (* equal contribution)

Preprint 2025

We introduce Reisom, a system for reconstructing in-scene object manipulation from a single monocular RGB video.

ZeroMimic: Distilling Robotic Manipulation Skills from Web Videos
ZeroMimic: Distilling Robotic Manipulation Skills from Web Videos

Junyao Shi*, Zhuolun Zhao*, Tianyou Wang, Ian Pedrozaโ€ , Amy Luoโ€ , Jie Wang, Jason Ma, Dinesh Jayaraman (* equal contribution, โ€  equal contribution)

ICRA 2025
๐Ÿ† Best Paper Award @ CVPR 2025 Workshop on 3D VLMs for Robotic Manipulation

We introduce ZeroMimic, a system that distills robotic manipulation skills from egocentric human web videos for zero-shot deployment in diverse environments with a variety of objects.

ZeroMimic: Distilling Robotic Manipulation Skills from Web Videos

Junyao Shi*, Zhuolun Zhao*, Tianyou Wang, Ian Pedrozaโ€ , Amy Luoโ€ , Jie Wang, Jason Ma, Dinesh Jayaraman (* equal contribution, โ€  equal contribution)

ICRA 2025
๐Ÿ† Best Paper Award @ CVPR 2025 Workshop on 3D VLMs for Robotic Manipulation

We introduce ZeroMimic, a system that distills robotic manipulation skills from egocentric human web videos for zero-shot deployment in diverse environments with a variety of objects.