Photo of Liwen Zhu
LZ

Liwen Zhu 朱丽雯

Algorithm Research Engineer

Tencent AI Lab (also as AI Platform), Beijing / Shenzhen

I am an Algorithm Research Engineer at Tencent AI Lab. My research interests span Reinforcement Learning, Large Language Model Agents, and Game AI.

Currently, I am working on the JueWu AI project, building RL agents for Honor of Kings that surpass professional players, and developing Companion AI deployed at 100M+ DAU scale. I also explore LLM-based strategic reasoning in complex game environments.

Previously, I worked on RLHF and LLM post-training at Tencent WXG. I received my M.S. from Peking University (advised by Prof. Zongqing Lu and Prof. Yonghong Tian).

liwenzhu [at] pku.edu.cn

Experience

Oct 2023 – Present
Tencent TEG / AI Lab (also as AI Platform), Shenzhen
Algorithm Research Engineer — RL, LLM Agent, Game AI
  • JueWu AI: PPO-based RL for Honor of Kings, surpassing KPL pros. Showcased in national competition.
  • Companion AI: Offline RL covering 80+ heroes, deployed at 100M+ DAU.
  • LLM Game Agent: Strategic reasoning and planning in complex strategy games.
Jul 2022 – Oct 2023
Tencent WXG / Pattern Recognition Center, Beijing
Algorithm Research Engineer — LLM Post-training, RLHF
  • Built Reward Model pipeline for WeLM with gap alignment algorithm.
  • Led PPO-based RLHF alignment across multiple SFT versions; reduced reward hacking.
Oct 2021 – Mar 2022
Microsoft Research Asia, Beijing  Star of Tomorrow
Research Intern — Deep RL, Cloud Computing
  • RL-based online job scheduling outperforming FIFO and SJF heuristics.
  • Multi-task auxiliary learning to augment state representations.
May 2021 – Jul 2022
Tencent, Beijing / Shenzhen
Algorithm Intern — RL, NLP
  • Modeled relation extraction as RL problem; entity relations in financial news.
  • Aspect-based sentiment analysis with Docker deployment on ES.
Jul 2019 – Jul 2022
Pengcheng Laboratory, Shenzhen
Algorithm Intern — Multi-Agent RL, Traffic Signal Control
  • Decentralized traffic signal control via meta-learning with intrinsic rewards.
  • Developed Shenzhen traffic simulation dataset and PengBo RL Platform.

Selected Publications

  1. Meta Variationally Intrinsic Motivated Reinforcement Learning for Decentralized Traffic Signal Control
    Liwen Zhu, Peixi Peng, Zongqing Lu, Yonghong Tian
    IEEE Transactions on Knowledge and Data Engineering (TKDE), 2022
    [Paper] [Code]
  2. Enhance Reasoning for Large Language Models in the Game Werewolf
    Shuang Wu*, Liwen Zhu*, Tao Yang, Shiwei Xu, Qiang Fu, Wei Yang, Haobo Fu
    arXiv 2024
    [Paper] [Code]
  3. MTLight: Efficient Multi-Task Reinforcement Learning for Traffic Signal Control
    Liwen Zhu, Peixi Peng, Zongqing Lu, Yonghong Tian
    ICLR 2022 Workshop (Gamification and Multiagent Solutions)
    [Paper] [Code]
  4. Multi-Agent Coordination via Multi-Level Communication
    Ziluo Ding, Zeyuan Liu, Zhirui Fang, Kefan Su, Liwen Zhu, Zongqing Lu
    NeurIPS 2024
    [Paper]
  5. From Unknown to Known: An AI Coaching Problem in Open-World Environments
    Xuejie Liu, Anji Liu, Zhengxinyue, Liwen Zhu, Zihao Wang, Xiaojuan Tang, Haowei Lin, Haobo Fu, Yitao Liang
    EMNLP 2024
    [Paper]
  6. An Advanced Reinforcement Learning Framework for Online Scheduling of Deferrable Workloads in Cloud Computing
    Liwen Zhu*, Hang Dong*, Zhao Shan, Bo Qiao, Yonghong Tian
    arXiv 2024
    [Paper]

Patents

Education

Sep 2019 – Jun 2022
Peking University, Beijing
M.S. in Computer Application Technology, GPA: 3.77 (Top 10%)
Advisor: Zongqing Lu, Yonghong Tian
Lab: Multimedia Learning Group, Institute of Digital Media (NELVT)
Sep 2015 – Jun 2019
Beijing University of Technology, Beijing
B.E. in Intelligent Traffic System, GPA: 3.91 (Top 1%)

Honors & Awards

2022Tencent Open Source Collaboration Award
2022Star of Tomorrow Honor, Microsoft Research Asia
2019Outstanding Graduates of Beijing
2019Beijing Excellent Undergraduate Thesis
2019Meritorious Winner, Mathematical Contest in Modeling (MCM/ICM)
2018National Scholarship, Ministry of Education of China
2018Grand Prize, iCAN International Innovation and Entrepreneurship Competition
2018Silver Award, “Chuang Qing Chun” Capital Entrepreneurship Competition

Academic Service