Hi! I’m a rising fourth-year undergrad at the University of Chicago studying Computer Science and Philosophy. I have been fortunate to work with Prof. Chenhao Tan at the Chicago Human+AI (CHAI) Lab, and with Prof. Alexandros G. Dimakis and Prof. Joseph E. Gonzalez at the UC Berkeley Sky Computing Lab. My research interests include agent benchmarking, agents for scientific discovery, and the self-improvement of agents. I’m currently doing an internship at Bespoke Labs, working on coding data.
Outside of research, I play the clarinet and have been part of the Chamber Music Program at UChicago since 2023. I also read in neuroscience and philosophy (I have been reading Levinas and Weil these days), and I am currently learning ancient Greek.
I’m always open to collaboration or just to chat about anything (mingxuanl@uchicago.edu)!
Publications
-
Mingxuan Li, Huanzhi Mao, Qiuyang Mang, Hanchen Li, Joseph E. Gonzalez, Alexandros G. Dimakis. EnvOpt: Automated Optimization for Agent Environments. Preprint.
[Paper] [Code] -
Mingxuan Li*, Huanzhi Mao*, Ruitao Zheng, Jarod Stanbury James, Yunze Li, Natanya Anderson, Joseph E. Gonzalez, Alexandros G. Dimakis. SlackBench: Benchmarking Agents on Collaborative Projects Grounded in Real Code Repositories. NeurIPS 2026 (Evaluations & Datasets Track).
[Paper] [Data] -
Mingxuan Li, Qirun Dai, Heran Wang, Chenhao Tan. Emergent Misalignment Is Not Magical. EMNLP 2026 (main).
[Paper] [Code] [Models & Data] -
Huanzhi Mao*, Mingxuan Li*, Ruoxi Wu, Shishir G. Patil, Joseph E. Gonzalez. Berkeley Function Calling Leaderboard (BFCL) V4. Agentic Evaluation — Part 3: Format Sensitivity for Tool Calls.
[Blog] [Code] -
Mingxuan Li, Hanchen Li, Chenhao Tan. HypoEval: Hypothesis-Guided Evaluation for Natural Language Generation. ACL 2026 (main).
[Paper] [Code] -
Haokun Liu*, Yangqiaoyu Zhou*, Mingxuan Li*, Chenfei Yuan, Chenhao Tan. Literature Meets Data: A Synergistic Approach to Hypothesis Generation. ACL 2025 (main).
[Paper] [Code]
* denotes equal contribution.