I am a visiting student at Stanford, advised by Professor Chelsea Finn. Currently, I am a senior undergraduate student in Computer Science at USC. I am also fortunate to work under the guidance of Dr. Jim Fan and Professor Yuke Zhu at NVIDIA Research. Previously, I worked with Professor Joseph J. Lim and Dr. Youngwoon Lee at USC.
I am broadly interested in robot learning. My research focus is on scalable methods that enable robots to do long-horizon tasks. I am particularly interested in building agents that can memorize and imagine – the core capabilities of human intelligence. My research relies on offline data and trains robots to absorb information from explicit demonstrations, general observation, and its own experimentation. Ultimately, I hope that robots and machines can equip with high-level cognitive skills to assist people and contribute to fundamental science, education, and space exploration.
I deeply believe in human ingenuity and the potential of AI. My goal for the next 20 years is to become a university professor and I want to establish the next Bell Labs — the idea factory that revolutionized the world for the better.
- [Mar, 2023] I am joining Chelsea Finn’s lab at Stanford as a visiting student
- [Feb, 2023] I gave a talk at MILA on skill-based and model-based RL
- [Sep, 2022] One paper is accepted to CoRL 2022
- [May, 2022] I received the USC Provost’s Research Fellowship
Skill-based Model-based Reinforcement Learning
Lucy Xiaoyang Shi, Joseph J. Lim, Youngwoon Lee
Conference on Robot Learning (CoRL), 2022
paper / project / code / twitter /
We devise a method that enables model-based RL on long-horizon, sparse-reward tasks, allowing us to learn with 5x less samples.
Within deep reinforcement learning, embodied AI, and robotics, I am particularly interested in:
Model-based learning: build internal world models of the embodied self and the world to develop a general understanding and improve sample-efficiency in learning through planning.
Hierarchical / Skill-based learning: enable long-horizon task learning through temporal abstraction and guided exploration.
Representation learning: autonomously infer and discover the underlying temporal, spatial, and causal structures of the input across different modalities for efficient decision-making and generalization.
A sample of other topics that I’m also very curious about: generative models, external memory, and learning from human videos.
- In-context Meta-curriculum Learning, advised by Dr. Jim Fan and Prof. Yuke Zhu, 2022-2023
- Skill-based Model-based RL, advised by Youngwoon Lee and Prof. Joseph Lim, 2021-2022
- Human Data Augmentation, advised by Prof. Stefanos Nikolaidis, 2022
- Unsupervised Exploration with Episodic Memory, advised by Siyuan Li and Prof. Chongjie Zhang, 2021
- Quantum Mechanics for Next-gen Spacecrafts, advised by Dr. Rahul Rughani and Prof. David Barnhart, 2020
- Avionics R&D in Rocketry, advised by Dr. Paul Giuliano, 2019-2021
- (Subscale) Nuclear Thermal Rocket Engine, advised by Dr. Charles Radovich and Dr. Matthew Gilpin, 2019-2020
A mountain of gratitude for people who have kindly mentored me and inspired me with their vision and passion!
Teaching and Mentoring
- CSCI 270: Introduction to Algorithms, Spring 2022 & Fall 2022
- First-generation student mentor, 2021
- Public Speaking tutor, 2020
In my free time, I enjoy reading, bouldering, and hiking. I also like exploring new places, primarily through casual runs or getting lost in the wild. I love living everyday as an adventure — and this is especially the case when I invite friends to taste the novel-flavored cookies I just baked. :)
Below I expand a little bit on some fun experiences that have shaped the way I think.
Envision a home on Mars
With my two brilliant friends Jessica Yuan and Albert Sun we published a paper exploring a Mars city state of 1 million people. We were the only undergraduate team to win the finalist (10 /176) and present at the 23rd Annual International Mars Society Convention. The paper was also later included as a chapter in the book Mars City States: New Societies for a New World.
Looking back, it was a wonderful summer of reading and thinking intensively to paint a coherent picture of the future. It gave me an epiphany that a myriad of things in the world are profoundly connected, and AI will be an important driven force behind many things I care about — like an interplanetary humanity, robotics, and universal education.
I once worked with a group of physicist and engineers to experimentally analyze the quantum theory quantized inertia as an alternative to dark matter. As far as I know the experiments haven’t been fruitful so far and we still have no proof / disproof for this quantum theory or even the dark matter. But I had a ton of fun learning about puzzles in quantum physics along the way! This experience also trained me to question the fundamental and think from first principles.
Literally “Rocket Science”
In parallel to the physics stuff I was also an avionics engineer at the USC Rocket Propulsion Lab (think: build brains for rockets). We are the first student group to successfully launch and recover an entirely student-designed rocket (Traveler IV) past the Karman line, the recognized boundary of space at 100 km (328,084 ft)! There were so many difficulties to overcome, but launching a rocket in desert is one of my best college memories! I love this group of people — they are ambitious, hungry for getting things to work, and incredibly caring. They also taught me if you deeply believe in something, you are halfway through achieving it. If you care about something deep enough, the unfavorable reality becomes negotiable.
A Quest for Meaning
I wasted a lot of misguided, youthful time thinking about the following three questions:
- Why does the universe exist?
- Why does humanity exist?
- Why do I exist?
I like reading philosophy, which humbles and inspires me (fun fact: I once majored in it!). So don’t be too surprised if we suddenly start discussing about the meaning of life in a casual conversation (I won’t make it scary, I promise). ;)
However, now instead of analyzing the limit of reasoning, I want to use my ability of reasoning to push the limit of reasoning. At present Artificial Intelligence is my best answer.