Compositional Visual Planning via Inference-Time Diffusion Scaling
International Conference on Learning Representations (ICLR), 2026
I am a Master’s student in Robotics at the Georgia Institute of Technology, working with Prof. Danfei Xu and Prof. Yongxin Chen. I maintain a GPA of 4.0/4.0 and expect to graduate in May 2026.
My research focuses on developing generative algorithms for long-horizon imitation learning and planning. I am particularly interested in compositional diffusion sampling methods that synthesize coherent long-horizon plans from short-horizon data. I also work on leveraging video-action models to improve policy learning through joint training with visual prediction and action generation.
Publications ( show selected / show by date / show by topic )
Topics: Decision Making
Experience
Robot Learning and Reasoning Lab
- EgoWAM: leveraging video-action models to improve policy learning through joint training with visual prediction and action generation.
- Compositional Visual Planning: introducing an inference-time compositional sampling approach that scales to unseen and long-horizon reasoning tasks.
School of Interactive Computing, Georgia Tech
- CS7643 Deep Learning. Supervisor: Prof. Danfei Xu.
- CS8803 Advanced Mobile Robotics. Supervisor: Prof. Frank Dellaert.
Selected Awards and Honors
- First-class Undergraduate Academic Scholarship.2021–2023
- National Scholarship.2022
Technical Skills
Programming Languages
Python, C/C++, JavaScript, HTML/CSS, MATLAB.
Machine Learning
PyTorch, multi-node multi-GPU training.
Robotics
ROS, Gazebo, PyBullet, MuJoCo, SAPIEN.
Real Robot Experience
