Mingqian Zheng 郑鸣谦

prof_pic.jpg

Hi there!

I am a Ph.D. student in LTI at Carnegie Mellon University, co-advised by Carolyn Rosé and Maarten Sap. My research explores the dynamics of communication between humans and Large Language Models (LLMs), as well as interactions among multiple LLMs. I study how human-LLM exchanges differ from human-human interactions, with the goal of optimizing these conversations for better human-AI collaboration. My work examines LLM safety through refusal strategies, multi-turn interaction risks, and the alignment between user expectations and model behavior. I also investigate multi-agent social simulations to uncover LLMs’ behavioral patterns and social reasoning in complex social contexts.

Previously, I completed my Master’s in Survey and Data Science at the University of Michigan, where I was advised by Yajuan Si. I conducted research on NLP with David Jurgens in the Blablablab and also worked with David Flood as a member of the HPACC team. Prior to that, I got my Bachelor’s at NYU Shanghai with double majors in Mathematics and Data Science (Computer Science Concentration), where I worked with Hongyi Wen on Recommendation Systems.

Education

Sep 2024 - Present
Ph.D. in Language and Information Technology
Advisors: Carolyn Rosé, Maarten Sap
Sep 2022 - May 2024
M.S. in Survey and Data Science (Data Science Track)
Advisors: David Jurgens, Yajuan Si
Sep 2018 - May 2022
B.S. in Mathematics and B.S. in Data Science (Computer Science Track)
Advisors: Shuyang Ling, Hongyi Wen, Olivier Marin

News

Oct 2026 XYBench received the Best Resource Paper Award at EMNLP 2026! 🏆
Aug 2026 Our paper XYBench has been accepted to EMNLP 2026 main conference!
Jul 2026 Our paper Useless but Safe? Benchmarking Utility Recovery with User Intent Clarification in Multi-Turn Conversations has been accepted to COLM 2026!
Apr 2026 Our paper Imperfectly Cooperative Human-AI Interactions: Comparing the Impacts of Human and AI Attributes in Simulated and User Studies has been accepted to Findings of ACL 2026!
Aug 2025 Our paper Let Them Down Easy! Contextual Effects of LLM Guardrails on User Perceptions and Preferences was mentioned in this Forbes article on AI welfare

Industry Experience

May 2026 - Present
Research Intern, Open Ecosystem Team
Safety alignment for general long-horizon agent training. Mentors: Maarten Sap, Pradeep Dasigi

Selected Publications

  1. EMNLP 2026
    Akhila Yerukola, Jena D. Hwang, Mingqian Zheng, Jenna Godsey, Hyunwoo Kim, Valentina Pyatkin, Jennifer Hu, and Maarten Sap
    In EMNLP, 2026
  2. COLM 2026
    Mingqian Zheng, Malia Morgan*, Liwei Jiang*, Carolyn Rose, and Maarten Sap
    In COLM, 2026
  3. ACL 2026 Findings
    Myke C. Cohen*, Mingqian Zheng*, Neel Bhandari*, Hsien-Te Kao, Xuhui Zhou, Daniel Nguyen, Laura Cassani, Maarten Sap, and Svitlana Volkova
    In Findings of ACL, 2026
  4. EMNLP 2025 Findings
    Mingqian Zheng, Wenjia Hu, Patrick Zhao, Motahhare Eslami, Jena D. Hwang, Faeze Brahman, Carolyn Rose, and Maarten Sap
    In Findings of EMNLP, 2025
  5. EMNLP 2025
    Jiarui Liu, Yueqi Song*, Yunze Xiao*, Mingqian Zheng*, Lindia Tjuatja, Jana Schaich Borg, Mona Diab, and Maarten Sap
    In EMNLP, 2025
  6. NAACL 2025
    Yinuo Xu*, Hong Chen*, Sushrita Rakshit*, Aparna Ananthasubramaniam*, Omkar Yadav*, Mingqian Zheng*, Michael Jiang*, Lechen Zhang*, Bowen Yi*, Kenan Alkiek*, Abraham Israeli*, Bangzhao Shu*, Hua Shen*, Jiaxin Pei*, Haotian Zhang*, Miriam Schirmer*, and David Jurgens
    In NAACL, 2025
  7. EMNLP 2024 Findings
    Mingqian Zheng, Jiaxin Pei, Lajanugen Logeswaran, Moontae Lee, and David Jurgens
    In Findings of EMNLP, 2024