Homepage

About Me

Hi, I am Mengyi Deng, a Ph.D. student in Data Science and Analytics at The Hong Kong University of Science and Technology (Guangzhou). I work with Prof. Wei Wang and Prof. Zhijiang Guo.

My research focuses on large language model agents, clarification mechanisms, uncertainty modeling, preference optimization, and agentic reinforcement learning for tool-using systems.

I am especially interested in building reliable agent systems that can reason, ask useful clarifying questions, and use tools effectively in complex environments.

I expect to graduate between June 2027 and June 2028, and I am actively seeking research and engineering job opportunities.

Research Interests

  • LLM Agents: skill transfer and generalization, multi-step tool use, task planning, clarification mechanisms, and uncertainty modeling.
  • LLM Alignment: preference optimization, RLHF, DPO, GRPO, multi-stage fine-tuning, and alignment algorithms.
  • Reinforcement Learning and Generative Models: Agentic RL, complex reasoning tasks, and Generative Flow Networks.

Education

  1. Ph.D. in Data Science and Analytics, HKUST(GZ)

    Information Hub. Supervisors: Prof. Wei Wang and Prof. Zhijiang Guo.

  2. Research Master's in Data Science and Analytics, HKUST(GZ)

    GPA: 4.09/4.3. Thesis: Graph Generative Flow Network for SAT Solvers.

  3. B.S. in Communication Engineering, South China Agricultural University

    GPA: 92.3/100. Outstanding Undergraduate Graduate.

Publications

Uncertainty-Aware Clarification in LLM Agents with Information Gain

Mengyi DENG, Zhiwei Li, Xin Li, Tingyu ZHU, Ying Zhao, Zhijiang Guo, Wei Wang

ICML 2026 Conference

DGPO: Beyond Pairwise Preferences with Directional Consistent Groupwise Optimization

Mengyi DENG, Zhiwei Li, Xin Li, Tingyu ZHU, Yihui Zhang, Duyi Pan, Yulan YUAN, Zhijiang Guo, Wei Wang

ACL 2026 Findings

When Inverse Data Outperforms: Exploring the Pitfalls of Mixed Data in Multi-Stage Fine-Tuning

Mengyi DENG, Xin Li, Tingyu ZHU, Zhicheng Yang, Zhijiang Guo, Wei Wang

EMNLP 2025 Findings

Academic Service

  • Conference reviewer: ICLR 2025, ACL 2026, ICML 2026, NeurIPS 2026.

Honors And Awards

  • National Scholarship.
  • Mathematical Contest in Modeling, Honorable Mention, Apr. 2020.
  • Guangdong Provincial Undergraduate Mathematical Contest in Modeling, Second Prize and Third Prize, Oct. 2019.

Interests

  • Badminton
  • Swimming
  • Hiking