Haotian Zhong
Undergraduate researcher at Tsinghua University · AI Safety · LLM Systems
Beijing, China
I am an undergraduate at Tsinghua University pursuing a dual-degree program in Mathematics and Physics (B.S.) + Software Engineering (B.Eng.). My research interests include AI safety, LLM agent security, content provenance, and efficient long-context language-model systems.
I currently work on defenses against indirect prompt injection in LLM agents with Prof. Yinpeng Dong, and on AI Multibit Robust Watermark for content attribution with Prof. Yuheng Bu at UC Santa Barbara. The watermarking manuscript is under submission to ARR; the prompt-injection defense project is in progress. Previously, I worked on LLM post-training for crystal-structure prediction with Prof. Yang Liu and conducted research on LLM agent security at Tencent Xuanwu Lab.
I will join the University of California, San Diego as a UCEAP exchange student for the Winter and Spring quarters of 2027 (two quarters).