🎓 About Me

I am Linxuan Du, a student interested in Large Language Models, LLM Reasoning, LLM Agents, and Natural Language Processing.

My research focuses on enhancing the ability of LLMs and agents to solve complex problems through reflection and memory.

Research Experience

Master Student at Soochow University   2025.09 - Present

  • Advised by Juntao Li and Xiaobo Liang.
  • Research topics include NLP, LLM agents, Reflection, Memory, and Reasoning.

📰 News

  • 2026.04: One paper was received in ACL 2026 MainConference

📝 Publications

* denotes equal contribution.

Paper Figure 1
Linxuan Du*, Guangquan Xue*, Xiaobo Liang, Qipeng Huang, Yuyang Ding, Xinyu Shi, Zhang Yijun, Ji Qi, Wenpeng Zhu, Juntao Li, Min Zhang

TL;DR: We propose the Tree-structured Advantage Estimation (TRAE) framework, which gives more fine-grained rewards to each reflection step in multi-turn reinforcement learning and helps solve the Echo Trap problem in multi-turn reflection.

🏆 Honors and Awards

  • LanqiaoCup Programming Competition - Second Prize (National Level, Individual Competition)
  • Computer Design Competition of China - Thrid Prize (National Level)
  • Core members of national-level college student innovation and entrepreneurship projects - Second Member (National Level)
  • Soochow University Innovation and Entrepreneurship Special Prize Scholarship

📖 Educations

  • 2025.09 - Present, Master Student, Artificial Intelligence Research Institute, Soochw University
  • 2021.09 - 2025.06, B.Eng., School of Computer Science and Technology, Soochow University