About
I am a Ph.D. candidate in Machine Learning at Georgia Tech, advised by Prof. Yongxin Chen and Prof. Humphrey Shi. My research focuses on world models and agents: generative models that understand, predict, and interact with the physical world, and the agents that plan and act with them.
Most recently I spent 2026 in the World Modeling team at Waymo, working on world-model pre-training, 3D Gaussian Splatting mid-training, and distillation for driving simulation, and 2025 at ByteDance on camera-controllable video generation.
Before Georgia Tech, I studied multi-object tracking with Prof. Jiaya Jia at CUHK, neural-operator control of turbulent flows with Prof. Anima Anandkumar at Caltech, neural-field representations with Prof. Junchi Yan at SJTU, program-guided agents with Prof. Le Song, and policy learning with Prof. Jiajun Wu and Prof. Josh Tenenbaum at MIT CSAIL. I received my B.S. in Computer Science from Shanghai Jiao Tong University (Zhiyuan Honors Program, ranked 3/145).
- world models
- agents
- video generation
- flow matching & diffusion
- neural fields
- 3D vision
Publications
All papers, newest first. Thumbnails are the teaser figure of each paper. A citation-sorted list is on Google Scholar.
2026
-
RMA: An Agentic System for Research-Level Mathematical Problems
LLA Workshop 2026
-
MV-S2V: Multi-View Subject-Consistent Video Generation
SIGGRAPH 2026
2025
2024
-
Grounding and Enhancing Grid-based Models for Neural Fields
CVPR 2024Best Paper Award CandidateOral, top 0.2%
2022
-
Tracking Objects as Pixel-wise Distributions
ECCV 2022Oral, top 4%
-
End-to-end View Synthesis via NeRF Attention
arXiv preprint 2022
2021
-
How to Design Sample and Computationally Efficient VQA Models
arXiv preprint 2021
2018
-
Estimating 6D Pose from Localizing Designated Surface Keypoints
arXiv preprint 2018
-
PointSIFT: A SIFT-like Network Module for 3D Point Cloud Semantic Segmentation
arXiv preprint 2018
Experience
Waymo, World Modeling · Software Engineer InternWorld-model pre-training, 3DGS mid-training, and distillation; action-conditioned world models from driving logs and counterfactual videos. Mountain View, CA.
ByteDance, E-commerce · Machine Learning Engineer InternCamera-controllable video generation (CETCam), human-object interaction videos (LOOM), multi-view subject-consistent video (MV-S2V). San Jose, CA.
Georgia Tech · Ph.D. in Machine LearningAdvisors: Prof. Yongxin Chen and Prof. Humphrey Shi. M.S. in Computer Science, 2026.
Caltech, Anima Lab · Research AssistantPhysics-informed neural-operator predictive control for turbulent drag reduction, with Prof. Anima Anandkumar.
Shanghai Jiao Tong University, ReThinkLab · Research AssistantGrid-based neural fields (MulFAGrid), with Prof. Junchi Yan and Dr. Fenglei Fan.
CUHK, Deep Vision Lab & SmartMore · Research StudentMulti-object tracking as pixel-wise distributions, with Prof. Jiaya Jia.
Georgia Tech, Machine Learning Group · Research AssistantProgram-guided agents (ProTo), with Prof. Le Song.
SenseTime Research, Fundamental Vision · Research InternMixed-precision training for object detection, with Dr. Jifeng Dai.
MIT CSAIL · Visiting StudentRoutines from a single demonstration for imitation and reinforcement learning, with Prof. Joshua B. Tenenbaum and Prof. Jiajun Wu.
Shanghai Jiao Tong University, MVIG · Research Assistant6D object pose estimation and point-cloud segmentation, with Prof. Cewu Lu.
Shanghai Jiao Tong University · B.S. in Computer Science and TechnologyZhiyuan Honors Engineering Program. GPA 91.90/100, ranked 3/145.
Software
- UnboundedNeRFPytorch Simple, fast unbounded and large-scale NeRFs; official code of the CVPR 2024 paper.1.3k stars
- ResearchMathAgent An agentic system for research-level mathematical problems.29 stars
- LapFlow Laplacian multi-scale flow matching, official code.7 stars
- betapose 6D pose estimation from designated surface keypoints.70 stars
Honors & service
Honors
- CVPR 2024 Best Paper Award Candidate and Oral Presentation
- ECCV 2022 Oral Presentation
- Extraordinary Bachelor's Thesis in Computer Science, SJTU, 2021
- National Scholarship (ranked 1/150), 2018
- First-class Academic Excellence Scholarship, SJTU (top 1%), 2018
- Meritorious Winner, Mathematical Contest in Modeling, 2018
- Shanghai Municipal Scholarship, 2017
- Zhiyuan Honors Scholarship, 2016 – 2020
- First Prize, Chinese Physics Olympiad, Shanxi Province, 2015
Reviewing
- Conferences: ICML, ICLR, NeurIPS, AAAI, CVPR, ICCV, ECCV, CICAI
- Journals: International Journal of Computer Vision, Neural Computing, Physics of Fluids
Teaching
- Teaching assistant, AE 3531 Control System Analysis and Design, Georgia Tech
- Teaching assistant, ENGG2440A Discrete Mathematics for Engineers, CUHK, 2021