Zehao Wen

zwen19 [at] jh [dot] edu

I'm a sophomore at Johns Hopkins pursuing B.S. in Computer Science, where I'm fortunate to be advised by Prof. Tianmin Shu and Prof. Homanga Bharadhwaj.

My current research interest lies in experience learning. How can we power both LLM and robotic agents to continuously learn and improve from their experiences?

I enjoy photography, collecting vinyls, and reading sci-fis. I love "chill walking": wandering without a goal but enjoying the movement, the talks, and the thoughts along.

News

May 2026

My blog post Flow Matching is out. It explains flow matching in a cohesive and intuitive way.

Publications

denotes equal contribution.

†

denotes equal mentorship.

AgentOdyssey: Open-Ended Long-Horizon Text Game Generation for Test-Time Continual Learning Agents

Zheyuan Zhang*, Zehao Wen*, Alvin Zhang, Andrew Wang, Jianwen Xie, Daniel Khashabi†, and Tianmin Shu†

COLM 2026 Workshop on Lifelong Agent

3D-Belief: Embodied Belief Inference via Generative 3D World Modeling

Yifan Yin*, Zehao Wen*, Jieneng Cheng, Zehan Zheng, Nanru Dai, Haojun Shi, Suyu Ye, Aydan Huang, Zheyuan Zhang, Alan Yuille, Jianwen Xie, Ayush Tewari, and Tianmin Shu

AnyHome: Open-Vocabulary Generation of Structured and Textured 3D Homes

Rao Fu*, Zehao Wen*, Zichen Liu*, and Srinath Sridhar

ECCV 2024

Interpolating Video-LLMs: Toward Longer-sequence LMMs in a Training-free Manner

Yuzhang Shang*, Bingxin Xu*, Weitai Kang, Mu Cai, Yuheng Li, Zehao Wen, Zhen Dong, Kurt Keutzer, Jae Lee Yong*, and Yan Yan*