Lirui Luo

PKU & JD TGT, Beijing, China

prof_pic_20260624.jpg

I am a Ph.D. student in the School of Intelligence at Peking University (PKU), majoring in Computer Science and Technology (Intelligent Science and Technology), advised by Cong Fang. I received my B.E. in Communication Engineering from the School of Electronic and Information Engineering at Beijing Jiaotong University (BJTU) in 2023.

I am currently an intern at JD.com’s Tech Genius Team (TGT, Top Young Technical Talent Program), working on Agentic RL post-training for e-commerce agents.

My research centers on continual reinforcement learning for agents. More specifically, I focus on three capabilities: (1) reinforcement learning, where an agent interacts with an environment, receives reward signals, and updates its parameters; (2) continual parameter updating, where an agent keeps updating its parameters over a stream of incoming tasks to master the full task stream; and (3) continual memory, where RL post-training teaches an agent to use tools to maintain and retrieve appropriate memory during long-horizon tasks.

news

Jun 24, 2026 Joined JD.com’s Tech Genius Team (TGT, Top Young Technical Talent Program) as an intern, working on Agentic RL post-training for e-commerce agents.
May 05, 2026 SPHERE accepted by ICML 2026.
Jan 30, 2026 MVR accepted by ICLR 2026.
Jun 13, 2024 INSIGHT has been selected as a spotlight-designated paper.
May 22, 2024 INSIGHT accepted by ICML 2024.

experiences

publications

  1. ICML
    sphere-teaser.png
    SPHERE: Mitigating the Loss of Spectral Plasticity in Mixture-of-Experts for Deep Reinforcement Learning
    Lirui Luo, Guoxi Zhang, Hongming Xu, and 2 more authors
    In Proceedings of the 43rd International Conference on Machine Learning , 2026
  2. ICLR
    mvr-teaser.png
    MVR: Multi-view Video Reward Shaping for Reinforcement Learning
    Lirui Luo, Guoxi Zhang, Hongming Xu, and 3 more authors
    ICLR, 2026
  3. ICML
    teaser-1.png
    End-to-End Neuro-Symbolic Reinforcement Learning with Textual Explanations
    Lirui Luo, Guoxi Zhang, Hongming Xu, and 3 more authors
    ICML, 2024