About me

I am Haomin Zhuang, a Ph.D. student in MINE Lab at the University of Notre Dame, advised by Prof. Xiangliang Zhang. I received my Bachelor’s degree in Software Engineering from South China University of Technology.

My research focuses on large language models (LLMs) — making them safer, more efficient, and better reasoners. I work across three interconnected directions: machine unlearning (selectively removing knowledge from LLMs, including in Mixture-of-Experts architectures [ACL 2025]), AI safety and alignment, and reasoning enhancement via reinforcement learning (e.g., multi-temperature strategies for RLVR [arXiv]).

Previously, I was a research intern at the OPTML Group (advised by Prof. Sijia Liu) and IntelliSys Lab (advised by Prof. Hao Wang). In summer 2025, I interned at ByteDance (TikTok), working on e-commerce chatbot systems.

Service: Reviewer for ICLR, ICML, NeurIPS, KDD, ACL, EMNLP, NAACL, TDSC, and TAI

News

• [Mar. 2026] Open-sourced AgentClick — a human-in-the-loop browser review layer for terminal agents, enabling inspect-and-edit before autonomous actions execute.
• [Oct. 2025] New preprint on arXiv: Exploring Multi-Temperature Strategies for Token- and Rollout-Level Control in RLVR
• [Jul. 2025] Paper accepted at ACL 2025: SEUF: Is Unlearning One Expert Enough for Mixture-of-Experts LLMs?