Luze Sun
(he/him)
PhD Student
Research interests
- AI security
Education
- MSE in Systems Engineering, University of Pennsylvania
- BSc in Computer Science, University of Bristol — UK
Biography
Luze Sun is a PhD student in the Khoury College of Computer Sciences at Northeastern University, based in Boston. He is advised by Alina Oprea.
With a passion for the safe use of large language models (LLMs), Sun enjoys framing misuse scenarios and stress-testing models to uncover findings that can be turned into reproducible benchmarks and usable safety tools. He was drawn to the subject during his time as a researcher at the University of Pennsylvania and ETH Zurich, when he saw standard single-turn detectors miss covert, multi-step attacks. While at those institutions, he developed covert-adversary benchmarks and stateful pipelines while also quantifying the “jailbreak tax” — the amount of LLM utility loss required to ensure a model’s safe use.
Sun joined Khoury College in 2025 to study LLM security, including the detection of misuse, safety/utility tradeoffs, scalable evaluation, and human-centered safety tools. He’s particularly interested in applying developments in human–computer interaction to LLM safety. Sun’s work has previously appeared in ICML and ICLR.