Pursuing Minimal Sufficiency in Spatial Reasoning
ICLR, 2026
A dual-agent framework that extracts and refines minimal sufficient 3D information for spatial reasoning.
Yejie GuoI am a first-year Ph.D. student in EECS at University of California, Merced, advised by Prof. Ming-Hsuan Yang. I received my B.Eng. in Artificial Intelligence (Honor Class) from Shanghai Jiao Tong University in 2026, where I was co-advised by Prof. Yong-Lu Li and Prof. Cewu Lu. During my undergraduate studies, I also spent a semester at EPFL as an exchange student. |
 |
Research
I am interested in vision-language models and intelligent agents. My research explores how AI systems perceive and reason about the visual world, with work spanning spatial reasoning, 3D scene editing, and visual perception.
Edit3r: Instant 3D Scene Editing from Sparse Unposed Images
ECCV, 2026 (Oral)
Feed-forward 3D scene editing from sparse unposed images, producing edited 3D Gaussians in a single pass.
Revisiting Articulated Parts Perception in Robot Manipulation
CVPR Findings, 2026
A scalable geometric representation and data collection framework for articulated-part perception and robotic manipulation.
M3-VOS: Multi-Phase, Multi-Transition, and Multi-Scenery Video Object Segmentation
CVPR, 2025
A benchmark and reversal-refinement method for segmenting objects through complex phase and appearance transitions.
* Equal contribution.
Education
University of California, MercedSep. 2026 – Present
PhD in EECS · Advisor: Prof. Ming-Hsuan Yang
Shanghai Jiao Tong UniversitySep. 2022 – Jun. 2026
B.Eng. in Artificial Intelligence (Honor Class)
École Polytechnique Fédérale de LausanneAug. 2024 – Jan. 2025
Semester exchange