Xun Xu
B.S. Student in Computer Science
Fudan University

About Me

I am a Computer Science undergraduate at Fudan University. My research interests center on self-evolving LLM agents, agent memory, and agentic reinforcement learning.

Most recently, I was a research intern at the NUS LV Lab, advised by Prof. Shuicheng Yan, where I studied how agents can learn from experience through coordinated policy learning and memory updates. Previously, I worked on latent memory for long-context reasoning at the FDU FVL Lab and on LLM agents as an LLM Agent Development Intern at Xiaohongshu (RedNote).

Education
  • Fudan University
    Fudan University
    Sep. 2023 - Jun. 2027
    School of Computer Science
    B.S. in Computer Science
Experience
  • NUS LV Lab
    NUS LV Lab
    Mar. 2026 - Jul. 2026
    Advised by Prof. Shuicheng Yan
    Research Intern · Self-Evolving Agents
  • Xiaohongshu (RedNote)
    Xiaohongshu (RedNote)
    Nov. 2025 - Feb. 2026
    LLM Agent Development Intern
  • FDU FVL Lab
    FDU FVL Lab
    Jun. 2025 - Sep. 2025
    Advised by Prof. Henghui Ding
    Research Intern · Latent Memory for LLMs
Honors & Awards
  • Outstanding Student Award with Scholarship
    2025–2026
News
2026
OPD-Evolver was accepted to EMNLP 2026.
Aug 22
Joined the NUS LV Lab as a research intern, working on self-evolving agents, agent memory, and agentic reinforcement learning.
Feb 28
2025
Joined Xiaohongshu (RedNote) as an LLM Agent Development Intern, working on LLM-based agents for cluster diagnosis.
Oct 31
Joined the FDU FVL Lab as a research intern to study latent memory for long-context LLM reasoning.
May 31
Selected Publications (view all )
Who Teaches Which Token? Verifier-Gated Multi-Expert On-Policy Distillation for Scientific Reasoning
Who Teaches Which Token? Verifier-Gated Multi-Expert On-Policy Distillation for Scientific Reasoning

Xun Xu, Zaixi Zhang

A multi-expert on-policy distillation framework for scientific reasoning that uses verifier feedback to decide which expert should teach each token. VG-OPD localizes and weights expert supervision along the student's reasoning trajectory, improving 4B and 8B students across seven benchmarks.

[Paper] [Code]

Who Teaches Which Token? Verifier-Gated Multi-Expert On-Policy Distillation for Scientific Reasoning

Xun Xu, Zaixi Zhang

A multi-expert on-policy distillation framework for scientific reasoning that uses verifier feedback to decide which expert should teach each token. VG-OPD localizes and weights expert supervision along the student's reasoning trajectory, improving 4B and 8B students across seven benchmarks.

[Paper] [Code]

OPD-Evolver: Cultivating Holistic Agent Evolver via On-Policy Distillation
OPD-Evolver: Cultivating Holistic Agent Evolver via On-Policy Distillation

Guibin Zhang*, Xun Xu*, Yanwei Yue, Zikun Su, Wangchunshu Zhou, Xiaobin Hu#, Shuicheng Yan# (* equal contribution, # corresponding author)

Conference on Empirical Methods in Natural Language Processing (EMNLP) 2026

A slow-fast co-evolution framework that teaches agents to improve through experience. OPD-Evolver combines a four-level memory hierarchy with outcome-calibrated attribution and privileged-hindsight distillation.

[Paper] [Code]

OPD-Evolver: Cultivating Holistic Agent Evolver via On-Policy Distillation

Guibin Zhang*, Xun Xu*, Yanwei Yue, Zikun Su, Wangchunshu Zhou, Xiaobin Hu#, Shuicheng Yan# (* equal contribution, # corresponding author)

Conference on Empirical Methods in Natural Language Processing (EMNLP) 2026

A slow-fast co-evolution framework that teaches agents to improve through experience. OPD-Evolver combines a four-level memory hierarchy with outcome-calibrated attribution and privileged-hindsight distillation.

[Paper] [Code]

All publications