Tianyuan Yuan (袁天远)

About Me

I am a fourth-year Ph.D. candidate at the Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University, advised by Prof. Hang Zhao. Before that, I received my Bachelor’s degree from Peking University.

My goal is to enable machines to perceive and interact with the physical world intelligently. To this end, my research bridges Computer Vision and End-to-End Learning, specifically within the contexts of Embodied AI and Autonomous Driving. Recently, I have been focusing on developing Vision-Language-Action (VLA) models and World-Action models to enhance robotic capabilities.

Research Interests

• Embodied AI: VLA Models, World-Action Models, Pretraining
• Autonomous Driving: End-to-End Learning, Mapping

News

• [Mar. 2026] We released Fast-WAM to answer the key question for WAMs: Do WAMs benefit mainly from explicit future imagination at inference, or from video modeling during training? We propose Fast-WAM, a World Action Model that removes explicit future imagination at test time.
• [Jan. 2026] Our paper DepthVLA and Galaxea Open-World Dataset is accepted by ICRA 2026.
• [Jan. 2026] We released G0Plus., our latest pre-trained VLA model for multi-task robot manipulation.
• [Sep. 2025] We released Galaxea Open-World Dataset and G0 VLA on Github and Huggingface.
• [Mar. 2025] Our paper LONG3R is accepted by ICCV 2025.

Publications (* indicates equal contribution)

Powered by Jekyll and Minimal Light theme.