01
3D Embodied Intelligence
3D representations for embodied agents, including neural rendering, Gaussian Splatting, scene reconstruction, and representations that support interaction and physical reasoning.
Embodied AI · 3D Vision · Robot Learning
Chen Liang 梁宸
I am an undergraduate student in the Zhiyuan John Class at Shanghai Jiao Tong University . From February to August 2026, I was a visiting student at UMass Amherst / MIT-IBM Watson AI Lab, working with Prof. Chuang Gan.
My research interests lie at the intersection of embodied intelligence, 3D representation learning, simulation, and world models. I am particularly interested in building scalable learning and evaluation pipelines that connect realistic 3D environments, physical simulation, and general-purpose robot policies.
Biography
I am a computer science undergraduate in the 2023 Zhiyuan John Class at Shanghai Jiao Tong University. In March 2025, I joined Prof. Yunbo Wang's group as a research intern, where I began working on world models, computer vision, and 3D representation learning.
From February to August 2026, I was a visiting student at UMass Amherst / MIT-IBM Watson AI Lab, working with Prof. Chuang Gan. My research there focused on embodied intelligence, 3D representations, simulation, and scalable robot learning.
I am particularly interested in how 3D scene representations and simulation can support scalable embodied learning: from reconstructing and generating interactive environments, to training and evaluating robot policies inside them, and ultimately transferring learned capabilities to the physical world.
More broadly, my interests include world models, vision-language-action models, Real2Sim/Sim2Real, and robot learning.
News
- Completed my visit to UMass Amherst / MIT-IBM Watson AI Lab, where I worked with Prof. Chuang Gan.
- Joined UMass Amherst / MIT-IBM Watson AI Lab as a visiting student, working with Prof. Chuang Gan.
- Our work EMGauss was accepted to CVPR 2026.
- Joined Prof. Yunbo Wang's group at Shanghai Jiao Tong University as a research intern.
Research Interests
02
Simulation & Real2Sim
Building scalable and realistic simulation pipelines for robot learning, policy evaluation, domain randomization, and closing the gap between simulated and real environments.
03
World Models & Robot Policies
World models, vision-language-action models, diffusion and flow-based robot policies, and learning representations of environment dynamics for planning and control.
04
Efficient Embodied Models
Efficient inference and deployment of embodied foundation models, including the interaction between model architecture, perception, policy inference, and hardware constraints.
Publications
Coming Soon
Simulation-Grounded Embodied Intelligence
Ongoing research on simulation-grounded robot learning, with a focus on improving the transfer, evaluation, and adaptation of embodied policies in real-world environments.
More details will be available as the project progresses.
Preprint
CVPR 2026
EMGauss: Continuous Slice-to-3D Reconstruction via Dynamic Gaussian Modeling in Volume Electron Microscopy
EMGauss reframes reconstruction from planar scanned 2D slices as dynamic 3D Gaussian rendering, enabling continuous slice synthesis without isotropy assumptions or large-scale pretraining.
Selected Projects
QEM Mesh Simplification
An implementation and evaluation of Garland–Heckbert Quadric Error Metrics for triangle-mesh simplification, with boundary preservation, flip detection, Metro evaluation, and comparisons against QSlim 2.0.
Education
Shanghai Jiao Tong University
Zhiyuan Honor College · John Hopcroft Class
B.S. in Computer Science
2023–Present
Shanghai High School
High School Diploma · Class 8
2020–2023
Research Experience
University of Massachusetts Amherst / MIT-IBM Watson AI Lab
Visiting Student · Working with Prof. Chuang Gan
Research on embodied intelligence, 3D representations, simulation, world models, and scalable robot-policy evaluation.
Feb. 2026–Aug. 2026
Shanghai Jiao Tong University
Research Intern · Prof. Yunbo Wang's Group
Research on world models, computer vision, and 3D representation learning.
Mar. 2025–Mar. 2026
Teaching & Service
Programming Course, Zhiyuan John Class
Teaching Assistant · Shanghai Jiao Tong University
Supported C++ programming assignments and prepared technical notes on concurrent data structures, build environments, and debugging.
SJTU
2023 Zhiyuan John Class
Class Representative
2023–Present
Miscellaneous
Outside research, I enjoy JRPGs, especially the Xenoblade Chronicles series, and light novels.
I also maintain a personal blog where I occasionally write notes on machine learning, 3D vision, topics I am studying, and JRPGs I enjoy.