I am Xiaolin (Will) Zhou, a PhD student in Information Systems at Arizona State University (ASU), working in the GLAD Lab, advised by Prof. Xiyang Hu.
I earned my M.S. in Computer Science at the University of Southern California (USC), advised by Prof. Ruishan Liu and Prof. Yue Zhao (FORTIS Lab).
My research focuses on LLM agents, reinforcement learning, and reliable AI. I study how agents behave when tools and environments deviate from their assumptions, and how training and evaluation can make these systems more dependable.
I am currently seeking summer 2027 research internships. Please get in touch.
NeurIPS 2026 · Evaluations and Datasets Track
RobustBench-TC evaluates tool-use agents under 22 types of disruption, while ToolRL-DR studies domain-randomized reinforcement learning for more reliable behavior.
arXiv preprint · Under review · 2026
A benchmark for agents that combine coding, graphical interaction, and visual feedback to repair software across web, game, mobile, and DevOps tasks, with deterministic tests of correctness.
arXiv preprint · Under review · 2026
A framework for comparing agent traces with and without skills, revealing behavioral changes that aggregate task success can miss.
arXiv preprint · Under review · 2026
An investigation of cross-language preferences in LLM judges and whether differences in fluency explain evaluation bias.
Journal of Physics: Conference Series, 2022
PhD in Information Systems
Fall 2026–present
M.S. in Computer Science
Jan 2024–Dec 2025
Program Committee Member, INFORMS Workshop on Data Science, 2026.
Reviewer, ACM Transactions on Intelligent Systems and Technology (TIST), 2026.