Yuncong Yang

I am currently a third-year PhD student at UMass Amherst, where I am supervised by Prof. Chuang Gan and collaborate with Prof. Yilun Du. Since June 2026, I have been a research intern at Meta MSL. Previously, I graduated from Columbia's Fu Foundation School of Engineering with both an M.S. and a B.S. in Computer Science (Summa Cum Laude). I was fortunate to work under the supervision of Prof. Shih-Fu Chang at Columbia University and with Dr. Jim Fan, Prof. Yuke Zhu, and Prof. Anima Anandkumar at NVIDIA Research.

My research interests lie in World Models, Embodied AI, and Robotics.

I am actively looking for Summer 2027 internship opportunities.

CV / Google Scholar / Twitter / Github

News

[2026.09] Two papers have been accepted to NeurIPS 2026, with SyncWorld selected for an Oral presentation!
[2026.06] I started my research internship at Meta MSL.
[2025.09] MindJourney is accepted by NeurIPS 2025!

Older news

[2025.03] 3D-Mem is accepted by CVPR 2025.
[2025.01] I will be interning at Microsoft Research this summer.
[2024.09] Starting my PhD journey at UMass Amherst!
[2023.01] One paper was accepted by ICLR 2023.
[2022.11] MineDojo won the Outstanding Paper Award at NeurIPS 2022!
[2022.07] One paper was accepted by ECCV 2022.
[2022.06] Officially released MineDojo!
[2022.01] Joined the NVIDIA AI research group, supervised by Dr. Jim Fan, Prof. Yuke Zhu, and Prof. Anima Anandkumar.
[2021.09] Joined the Digital Video and Multimedia (DVMM) Lab at Columbia University, supervised by Prof. Shih-Fu Chang.

Research

(* indicates equal contribution)

	SyncWorld: Visual Calibration Enables World Models as Zero-Shot Simulators Yuncong Yang, Zhengtao Han, Furkan Özyurt, Zeyuan Yang, Han Yang, Junyi Cao, Haoyu Zhen, Yilun Du, Chuang Gan NeurIPS 2026* (Oral) project page / paper / code / model We introduce SyncWorld, a visually calibrated world model that predicts the visual outcomes of numerical robot controls in unseen cameras, environments, and embodiments without downstream training.
	Action Images: End-to-End Policy Learning via Multiview Video Generation Haoyu Zhen, Zixian Gao, Qiao Sun, Yilin Zhao, Yuncong Yang, Yilun Du, Pengsheng Guo, Tsun-Hsuan Wang, Yi-Ling Qiao, Chuang Gan NeurIPS 2026 project page / paper / code We introduce Action Images, an end-to-end framework that learns robot policies through multiview video generation.
	Fast Spatial Memory with Elastic Test-Time Training Ziqiao Ma, Xueyang Yu, Haoyu Zhen, Yuncong Yang, Joyce Chai, Chuang Gan ECCV 2026 project page / paper / code / blog We introduce an elastic test-time training approach for fast, adaptive spatial memory in embodied agents.
	MindJourney: Test-Time Scaling with World Models for Spatial Reasoning Yuncong Yang, Jiageng Liu, Zheyuan Zhang, Siyuan Zhou, Reuben Tan, Jianwei Yang, Yilun Du, Chuang Gan NeurIPS 2025 project page / paper / code We proposed MindJourney, a test-time scaling framework that improves spatial reasoning with the assistance of controllable world models.
	Learning 3D Persistent Embodied World Models Siyuan Zhou, Yilun Du, Yuncong Yang, Lei Han, Peihao Chen, Dit-Yan Yeung, Chuang Gan NeurIPS 2025 paper / code
	3D-Mem: 3D Scene Memory for Embodied Exploration and Reasoning Yuncong Yang, Han Yang, Jiachen Zhou, Peihao Chen, Hongxin Zhang, Yilun Du, Chuang Gan CVPR 2025 project page / paper / code / twitter We proposed 3D-Mem, a framework that serves as 3D scene memory to empower embodied agents with lifelong exploration and reasoning abilities in 3D environments.
	TempCLR: Temporal Alignment Representation with Contrastive Learning Yuncong Yang, Jiawei Ma, Shiyuan Huang, Long Chen, Xudong Lin, Guangxing Han, Shih-Fu Chang ICLR 2023 paper / code We proposed TempCLR, a new contrastive learning framework that considers sequence-level temporal order consistency in Long-Video Understanding.
	MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge Linxi Fan, Guanzhi Wang, Yunfan Jiang, Ajay Mandlekar, Yuncong Yang, Haoyi Zhu, Andrew Tang, De-An Huang, Yuke Zhu, Animashree Anandkumar NeurIPS 2022 Datasets and Benchmarks Track (Outstanding Paper Award, Featured Paper Presentation) project page / paper / code We introduce MineDojo, a new framework based on the popular Minecraft game for building generally capable, open-ended embodied agents.
	Few-Shot End-to-End Object Detection via Constantly Concentrated Encoding across Heads Jiawei Ma, Guangxing Han, Shiyuan Huang, Yuncong Yang, Shih-Fu Chang ECCV 2022 paper

Teaching

Teaching Assistant: COMS 4732 Computer Vision II (Spring 2022)

Thanks for the template from Jon Barron!