Zisu Huang

M.S. Student at Fudan NLP Group · huangzs25@m.fudan.edu.cn

zisu.jpg

I am a first-year M.S. student in Computer Science at Fudan University, and a member of Fudan NLP Group, advised by A.P. Xiaoqing Zheng and Prof. Xuanjing Huang. I received my B.E. in Software Engineering from Fudan University in 2025.

I am currently a research intern in the Visual Computing Group at Microsoft Research Asia, working with Yifan Yang.

My research interests center on AI agents, with a particular focus on:

  • Personalized agents: developing more effective personalized agents through both model-level adaptation and harness/system design.
  • Agent evolution:
    • Parametric evolution: improving agentic capabilities through parameter updates, e.g., agentic RL.
    • Non-parametric evolution: expanding the capability boundary of agents in a training-free manner, e.g., through agent skills, orchestration, and harness design.

Feel free to reach out if you’d like to chat about research or collaborate.

News

May 22, 2026 Released two new works on agent skills: SkillOpt ( code stars ) and SkillLens ( code stars ). SkillOpt has sparked a lot of discussion in the community — thanks for the support! ✨
May 01, 2026 Our paper TRIP-Bench has been accepted to ICML 2026. 🎉
Apr 16, 2026 Check out our latest survey on reward hacking: Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges! 🔥
Apr 07, 2026 Our paper SteeM has been accepted to ACL 2026 (Main Conference). 🎊
Mar 03, 2026 Started a research internship at the Visual Computing Group, Microsoft Research Asia (Shanghai), working with Yifan Yang. 🎉
Jan 26, 2026 Our paper RECAST has been accepted to ICLR 2026. 🎊
Aug 21, 2025 Two papers accepted to EMNLP 2025 — SATER and FedQSN. 🎉

Survey

  1. Survey
    reward-hacking-survey.png
    Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges
    Xiaohua Wang*, Muzhao Tian*, Yuqi Zeng*, Zisu Huang*, and 19 more authors
    arXiv preprint, 2026
    Paper

Selected Publications (* indicates equal contribution)

See the full publication list →

  1. arXiv
    skillopt-teaser.png
    SkillOpt: Executive Strategy for Self-Evolving Agent Skills
    Yifan Yang*, Ziyang Gong*, Weiquan Huang*, Qihao Yang*, Ziwei Zhou*, Zisu Huang*, and 9 more authors
    arXiv preprint, 2026
    Paper Video Code Website
    3k+ stars in week 1 1M+ views🤗 #1 Paper of the Day
  2. arXiv
    skilllens-main.png
    From Raw Experience to Skill Consumption: A Systematic Study of Model-Generated Agent Skills
    Zisu Huang*, Jingwen Xu*, Yifan Yang, and 13 more authors
    arXiv preprint, 2026
    Paper Code Website
  3. ACL 2026
    steem.png
    Controllable Memory Usage: Balancing Anchoring and Innovation in Long-Term Human-Agent Interaction
    Zisu Huang*, Muzhao Tian*, Xiaohua Wang, and 8 more authors
    Annual Meeting of the Association for Computational Linguistics (ACL), 2026
    Main Conference
  4. ICML 2026
    trip-bench.png
    TRIP-Bench: A Benchmark for Long-Horizon Interactive Agents in Real-World Scenarios
    Yuanzhe Shen*, Zisu Huang*, Zhengyuan Wang*, and 14 more authors
    International Conference on Machine Learning (ICML), 2026
    Poster
  5. arXiv
    refiner.png
    Enhancing the Capability and Robustness of Large Language Models through Reinforcement Learning-Driven Query Refinement
    Zisu Huang*, Xiaohua Wang*, Feiran Zhang, and 5 more authors
    arXiv preprint, 2024
  6. ICLR 2026
    recast.png
    RECAST: Strengthening LLMs’ Complex Instruction Following with Constraint-Verifiable Data
    Wenhao Liu, Zhengkang Guo, Mingchen Xie, Jingwen Xu, Zisu Huang, and 9 more authors
    International Conference on Learning Representations (ICLR), 2026
    Poster

Miscellaneous

Recently, I have really enjoyed playing tennis 🎾 and have been working hard to improve my game. My favorite player is Novak Djokovic.