Yu Yuan

My name is Yu Yuan (袁煜), a final-year PhD student at Purdue ECE, under the supervision of Prof. Stanley H. Chan. Prior to this, I obtained my master and bachelor degree from Shanghai Jiao Tong University in 2023 and 2020, respectively. I am currently a research intern at Roblox Foundation AI, working on video world models. I interned with Adobe Research Multimodal Foundation Models team in summer 2026.

I aim to build physics-consistent and controllable generative worlds. I want to narrow the gap between generative models and our physical world. My approach is to identify meaningful problems, understand their underlying causes, and develop simple, effective solutions. My research explores how generative worlds are perceived through cameras Generative Photography [CVPR 2025 Highlight], how dynamics evolve over time NewtonGen [ICLR 2026], how spatiotemporal relationships are represented SeeU [CVPR 2026], and how different actions lead to different outcomes OptiWorld [NeurIPS 2026], and RVD.

I am seeking full-time Research Scientist / Member of Technical Staff opportunities in Video Generation and Physical World Models, with graduation expected in late 2026 or early 2027. Feel free to reach out!

Email / CV / Github / Scholar / LinkedIn / Flickr / Blog / Slides

(Generated by Nano Banana)

News

[2026/09] Released our Adobe internship project, RVD: a novel framework for video dynamics editing, training-free retiming, and appearance re-rendering.
[2026/09] OptiWorld has been accepted by NeurIPS 2026!
[2026/09] I have joined in Roblox Foundation AI for internship.
[2026/06] We released OptiWorld: which brings optimal control into video world generation.
[2026/05] I started my first internship at Adobe Research/Firefly Multimodal Foundation Models Team.
[2026/03] I passed my PhD Prelim Exam.
[2026/02] SeeU has been accepted by CVPR 2026!
[2026/01] NewtonGen has been accepted by ICLR 2026!
[2025/11] We released SeeU: a 4D understanding and video generation pipeline that reveals unseen time and space.
[2025/10] Our proposal for the 1st Generative AI for Photography (GAIP) Workshop has been accepted by WACV 2026.
[2025/09] We released NewtonGen: a physics-consistent and controllable text-to-video framework.

Talks

[2026/03] Toward Physics-Consistent and Controllable Generative Worlds @ SixteenMiles Labs .
[2026/03] Toward Physics-Consistent and Controllable Generative Worlds @ SFU VCR/AI Seminars .
[2025/12] Physics-consistent and controllable video generation @ ResearchTrend .
[2025/07] Generative photography @ VALSE .

Research

Generative Models

	Reimagine Video Dynamics Yu Yuan, Yawen Lu, Guoxian Song, Kevin Duarte, Ratheesh Kalarot, Di Chang, Xijun Wang, Stanley H. Chan Under Review proj / arXiv / code (soon)
	OptiWorld: Optimal Control for Video World Generation under Physical Constraints Yu Yuan, Jianhao Yuan, Xijun Wang, Daiqing Li, Liu He, Lu Ling, Stanley H. Chan NeurIPS 2026 proj / arXiv / dataset (soon) / code (soon)
	Beyond Visual Quality: A Study of Test-Time Planning with World Action Models Jianhao Yuan, Yu Yuan, Benjamin Ramtoula, Lukas Vierling, Paul Newman, Lars Kunze, Philip Torr, Daniele De Martini Under Review arXiv
	SeeU: Seeing the Unseen World via 4D Dynamics-aware Generation Yu Yuan, Tharindu Wickremasinghe, Zeeshan Nadir, Xijun Wang, Yiheng Chi, Stanley H. Chan CVPR 2026 proj / arXiv / dataset / code
	NewtonGen: Physics-Consistent and Controllable Text-to-Video Generation via Neural Newtonian Dynamics Yu Yuan, Xijun Wang, Tharindu Wickremasinghe, Zeeshan Nadir, Bole Ma, Stanley H. Chan ICLR 2026 proj / arXiv / code
	Generative Photography: Scene-Consistent Camera Control for Realistic Text-to-Image Synthesis Yu Yuan, Xijun Wang, Yichen Sheng, Prateek Chennuri, Xingguang Zhang, Stanley H. Chan CVPR 2025 Highlight & Demo proj / arxiv / dataset / demo / code
	Learning to Kindle the Starlight Yu Yuan, Jiaqi Wu, Lindong Wang, Zhongliang Jing, Henry Leung, Shuyuan Zhu, Han Pan arXiv 2023 arXiv / dataset / code

Computational Photography

	iHDR: Iterative HDR Imaging with Arbitrary Number of Inputs Yu Yuan, Yiheng Chi, Xingguang Zhang, Stanley H. Chan ICIP 2025 arXiv
	Astrophotography Turbulence Mitigation via Generative Models Joonyeoup Kim, Yu Yuan, Xingguang Zhang, Xijun Wang, Stanley H. Chan ICIP 2025 proj / arXiv

Real-world Vision

	Learning Phase Distortion with Selective State Space Models for Video Turbulence Mitigation Xingguang Zhang, Nicholas Chimitt, Xijun Wang, Yu Yuan, Stanley H. Chan CVPR 2025 Highlight proj / arXiv / code
	Restore What Matters: Lessons from Joint Restoration and Recognition Lanqing Guo¹, Xijun Wang¹, Minchul Kim¹, Yu Yuan, Wes Robbins, Xingguang Zhang, Stanley H. Chan, Zhangyang Wang, Xiaoming Liu Under Review
	Personalized Generative Low-light Image Denoising and Enhancement Xijun Wang, Prateek Chennuri, Yu Yuan, Bole Ma, Xingguang Zhang, Stanley H. Chan Under Review proj / arXiv
	RCMixer: Radar-camera Fusion Based on Vision Transformer for Robust Object Detection Lindong Wang, Hongya Tuo, Yu Yuan, Henry Leung, Zhongliang Jing JVCI 2024 doi
	Multimodal Image Fusion based on Hybrid CNN-Transformer and Non-local Cross-modal Attention Yu Yuan, Jiaqi Wu, Zhongliang Jing, Henry Leung, Han Pan arXiv 2023 arXiv / code

Leading Projects

Physical World Models (2026)
4D World Understanding and Dynamics Modeling (2025)
Physics-driven Video Generation (2025)
Generative Astrophotography Image Restoration (2024)
Video Restoration Classifier (2024)
An Underwater Binocular Multi-exposure Imaging System (2022)
Ultra-wideband-based Cooperative Multi-UAV Positioning Technology (2021)
A Modular Custom Aviation Model Design for Tyros (2020)

Patents

Learning a Compact Dynamics Token for Precise Video Motion Reconstruction and Editing. [Under Review].
A Data Pipeline for Generating Counterfactual Video Pairs from Real Videos. [Under Review].
Trajectory resolving and registering method based on autonomous underwater robot inertial navigation and ultra-short baseline positioning sensor. CN 114993313 A.

Misc.

Angler: Awarded "Fish of the Year" for catching the largest Goldeye (2023) and Shovelnose Sturgeon (2025) in Indiana
Landscape Photographer: One Milky Way photo was accepted to Astronomy Picture of the Day in Nightchina (2021)
Mountaineer: Arrived at the 2nd Peak of Zuimi MT. Siguniang (5276m, 2020)
Aviation Enthusiast: Experienced more than 40 types of manned aircraft, with 9 years of drone flying experience. I am a certified remote pilot under FAA Part 107 and plan to begin training for a private pilot license before graduation.

Last updated: