cv

PhD Student — Resilient Cyber-Physical Systems Lab

University of California, Irvine

Research on AI alignment and safety under Prof. Yasser Shoukry. Developing adversarial detection methods with formal guarantees, progress-monitored verification for VLA models, and LLM preference personalization.

• KoALA: Adversarial detector with provable guarantees (ICLR'26 submission)
• Progress-monitored verification for vision-language-action robot policies
• Reinforcement fine-tuning for LLM preference following

Visiting Student Researcher

California Institute of Technology

Research on human-robot collaboration under Prof. John Doyle. Combined large language models with model-predictive control for task planning and trajectory optimization.

• Human natural language to robotic skills and control
• LLM + MPC integration for robot task planning
• Visual-language model feedback for improved performance

Ph.D.

University of California, Irvine

Computer Engineering

M.Sc.

Karlsruhe Institute of Technology

Mechatronics and Information Technology (Robotics Track)

B.Eng.

Tongji University

Vehicle Engineering

Graduate Dean's Recruitment Fellowship

University of California, Irvine

Awarded to top 2% of incoming graduate students.

AGS Conference Stipend

University of California, Irvine

Awarded to top 5% of graduate students for conference travel.

China Automotive Science and Technology Award

China Society of Automotive Engineers

Highest prize for cloud control energy-saving algorithm development.

KoALA: KL-L0 Adversarial Detector via Label Agreement

arXiv (under review at ICLR 2026)

Adversarial detection method combining KL divergence with L0-based label agreement, with formal detection guarantees.

Optimizing Rare Word Accuracy in Direct Speech Translation with a Retrieval-and-Demonstration Approach

EMNLP 2024

Retrieval-augmented approach for rare word translation in end-to-end speech translation, achieving 17.6% improvement.