About Me
Tuen-Yue Tsui AI Researcher · University of Pennsylvania

I just do things I find I'm interested in, though I do have a preference on thinking across different fields.
I'm advised by Prof. Lingjie Liu, Prof. Konrad Kording, and Prof. Jiatao Gu.
Reinforcement Learning · Generative Models · Anything I Find Interesting
Research
Minimal Witness Reinforcement Learning
MWRL credits each proposal for the coverage the group's union of certified sets would lose without it. The credit, derived from the problem definition, asks for minimality and for every alternative at once, and...
Noise-Level Adjacency in Diffusion Training
The Lattice of Transition Laws
Mori–Zwanzig Formulation of Bayesian Filtering
Correct, minimal, and all
Many tasks have more than one minimal answer, none containing another. A reward that scores each rollout on its own cannot tell a new answer from a redundant copy of one it already has.
Projects
Vectorized Environment and Multi‑Agent RL for Teamfight Tactics
A vectorized simulator of Teamfight Tactics in Rust, exposed to Python through PyO3 and checked bit for bit against a Python reference...
Apodexis: Mathematical Proof‑Chain Visualization Tool
Research tool and formal-methods project. Built a native SwiftUI macOS app for organizing long mathematical and scientific proof chains as typed dependency...
Scalable Quadruped Imitation from Monocular Video
Retargeted quadruped motions from monocular videos via a learnable skeleton, removing the need for motion capture or manual keypoints.
Minimum-Snap Trajectory Generation for Quadrotors
Integrated VIO, an SE(3) controller and a ray-casting A* planner (30× faster, from 22 s to 0.7 s; first on the course...