I am a Computer Science undergraduate at HKU and an exchange student at Princeton. My research goal is building reliable, learning-capable AI agents through rigorous evaluation, verification, and interaction. My research focuses on AI agents, AI evaluation, computer-use agents, AI for mathematics, and reinforcement learning.
At Princeton, I work with Professor Pramod Viswanath, mentored by Zerui Cheng. At HKU, I work with Professor Tao Yu in the XLang Lab.
I am currently looking for PhD opportunities beginning in Fall 2027.
When Wrong Answers Look Right: Multi-Agent Debate for High-Precision Verification of Hard Competition Mathematics
AEGIS-D uses five role-specialized agents to cross-check competition-math solutions through structured debate and positive-evidence voting. On 195 hard-to-verify variants, it raises precision from 55.1% to 92.5% and reduces false positives by 8.3×.
Benchmarking Computer-Use Agents on Long-Horizon Real-World Tasks
OSWorld 2.0 evaluates agents on 108 realistic, long-horizon, end-to-end workflows across web and desktop applications. Each task requires sustained execution, cross-application state tracking, hidden-state recovery, and reliable final-state verification.
Open-Ended Computer Science Challenges for AI Agents
OpenFrontierCS is a community challenge for improving executable solvers on eight open-ended computer-science problems. Deterministic hidden-test evaluators provide partial scores, while promoted solutions become shared baselines for further improvement.
DASW replaces DDPO's uniform timestep weighting with budget-preserving weights from feature-space spread and scheduler variance. Experiments show stronger reward-diversity trade-offs on aesthetic and prompt-image alignment tasks.
B.Eng. in Computer Science
Selected coursework: artificial intelligence, machine learning, robotics, optimization, and numerical methods.
Undergraduate Exchange in Computer Science
Selected coursework: natural language processing, reinforcement learning, and deep learning.
Comate Team Intern
Teaching Assistant
Led weekly tutorials and supported students across three programming courses: ENGG1330, COMP2113, and COMP1117.
RMB 320,000 scholarship.
Sole recipient selected across HKU.
Recognized for academic achievement at the University of Hong Kong.
ICML 2026 AI4Math Workshop.
IELTS overall score of 8.0, including 8.0 in Speaking.
Former Mathematics Olympiad competitor.
Member of the HKU Rowing Team.
Always planning the next trip ✈️, happiest in the water 🚣 🏊, and usually up for cooking, singing, dancing, or a good stand-up set.