About
I am an undergraduate student in Artificial Intelligence at Xiamen University, advised by Fei Chao.
I am broadly interested in robot learning and embodied AI. My research focuses on how pretrained robot policies can be steered, adapted, and improved through interaction, and how these capabilities can translate into reliable behavior in the physical world.
In the long run, I hope to develop embodied agents that can not only acquire broad capabilities from large-scale pretraining, but also refine them through feedback and experience — adapting to new situations, recovering from failures, and continuing to improve after deployment.
News
[Sep, 2026] Joined the Embodied Robotics course at Xiamen University as a Teaching Assistant, contributing to hands-on laboratory design and technical instruction.
[Jun, 2026] PhysReflect-VLA was accepted to IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2026).
Selected Research
PhysReflect-VLA: Physical Feasibility and Self-Reflective Regulation for Reliable Vision-Language-Action Policies
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2026
PhysReflect-VLA improves long-horizon VLA control through physical feasibility evaluation and self-reflective failure recovery.
PAMAE: Phase-Aware-MoE Action Experts Towards Reliable Flow-Matching Vision-Language-Action Policies
arXiv preprint
PAMAE introduces phase-aware mixture-of-experts action generation for more reliable flow-matching VLA policies in multi-stage manipulation.
Research Interests
Within robot learning and embodied AI, I am particularly interested in:
Steering and post-training: adapt and refine pretrained robot policies through reinforcement learning, feedback, and inference-time intervention, turning broad pretrained capabilities into controllable and effective behavior.
Interactive robot learning: enable robots to continue improving through experience, using interaction with the physical world to adapt to new tasks, environments, and failures.
Reliable long-horizon behavior: develop robot policies that remain physically grounded and coherent over extended tasks, with the ability to detect errors, recover from failures, and adapt their behavior as execution unfolds.
I am also curious about predictive world models, continual learning, and multi-agent interaction as possible mechanisms for building more adaptive embodied agents.
Teaching and Mentoring
Embodied Robotics, Xiamen University — Teaching Assistant, Fall 2026
Contributing to hands-on laboratory design and technical instruction in robot control, imitation learning, and Vision-Language-Action models.
Misc
Outside research, I enjoy reading, wandering through cities, playing chess, and thinking about food that I may or may not ever cook. During my first year of college, I often spent weekends exploring cities without a fixed destination — streets, stations, bookstores, neighborhoods, and whatever happened to be along the way.
I also have a soft spot for old books, from Chinese classical literature to medieval writings, and for being a bit of a “cloud foodie” whenever real cooking is temporarily unavailable.
Chess is one of those interests I keep returning to. I occasionally spend time with a small group of fellow enthusiasts who enjoy building and experimenting with their own chess engines.
Much of it is delightfully informal: strange moves, long games, broken ideas, unexpected successes, and plenty of complete nonsense. Every now and then, ideas from machine learning sneak in, but I mostly enjoy it as a chance to play with models and exchange half-formed ideas with friends.
Not every curiosity needs a problem statement, a benchmark, or a conclusion. Some are worth following simply because they are interesting.
A Final Note
I still do not know exactly where research will take me, and I hope that remains true for a while. Some of the questions I care about today will probably disappear; others have not occurred to me yet.
What I hope to keep is the curiosity that made all of this interesting in the first place — the pleasure of finding something I do not understand, playing with it for a while, and occasionally watching an idea turn into something real.
