Binjie Zhang

Ph.D. Candidate, Show Lab, National University of Singapore.

prof_pic.jpg

Email: binjie97 [at] u.nus.edu

Show Lab, NUS

Singapore

I am a Ph.D. candidate in Computer Science at the National University of Singapore (NUS), advised by Assistant Prof. Mike Zheng Shou in Show Lab.

My research interests are LLM agents — harness design, memory, tool use, and self-evolution — together with vision–language and video understanding, world / predictive models, and lifelong and compatible representation learning. Broadly, I want to build systems where multi-modal agents can reuse what they already know, anticipate what comes next, and keep improving after deployment.

I am currently an AI Research Intern at TikTok (ByteDance) in Singapore, working end-to-end on LLM-agent systems — spanning agent harness design, self-evolving training loops, and long-horizon memory — and taking them from research prototype to production.

Before NUS I received my M.Eng. in Computer Science and Technology from Tsinghua University, supervised by Prof. Chun Yuan, and my B.Eng. in Information Engineering from East China University of Science and Technology (ECUST), where I ranked 1 / 92.

Before that, I spent several years at Tencent ARC Lab (Shenzhen) as an AI research intern, where I led a line of first-author work on backward-compatible representation learning and hot-refresh model upgrades for large-scale image and video retrieval (ICLR’22, IJCAI’22 long oral, AAAI’23, arXiv’23), as well as on cross-modality video understanding.

news

Mar 07, 2026 Submitted ReGRPO (reflection-augmented RL for tool-using agents) to ECCV 2026 as first author.
Jan 30, 2026 Two first-author papers submitted to ICML 2026: RCFC (lifelong imitation learning) and Ego-centric Predictive Model Conditioned on Hand Trajectories (top 15%).
Jan 06, 2026 Started as an AI Research Intern at TikTok (ByteDance) in Singapore, working end-to-end on LLM agents — harness, memory, tool use, and self-evolving training loops.
Feb 01, 2023 Darwinian Model Upgrades accepted to AAAI 2023.
Jan 10, 2023 Started as a Ph.D. student at Show Lab, NUS, advised by Prof. Mike Zheng Shou.

selected publications

  1. ECCV’26*
    ReGRPO: Reflection-Augmented Policy Optimization for Tool-Using Agents
    Binjie Zhang* and others
    In European Conference on Computer Vision (ECCV), 2026
    Under review
  2. ICML’26*
    prcfc.png
    RCFC: Lifelong Imitation Learning via Prototype Replay and Coarse-to-Fine Compatibility
    Binjie Zhang* and others
    In International Conference on Machine Learning (ICML), 2026
    Under review
  3. ICML’26*
    egopm.png
    Ego-centric Predictive Model Conditioned on Hand Trajectories
    Binjie Zhang* and others
    In International Conference on Machine Learning (ICML), 2026
    Under review — top 15%
    HTML
  4. arXiv’23*
    taca.png
    TaCA: Upgrading Your Visual Foundation Model with a Task-Agnostic Compatible Adapter
    Binjie Zhang*, Yixiao Ge, Xuyuan Xu, and 1 more author
    arXiv preprint arXiv:2306.12642, 2023
    HTML
  5. AAAI’23*
    Darwinian Model Upgrades: Model Evolving with Selective Compatibility
    Binjie Zhang*, Yixiao Ge, Xuyuan Xu, and 1 more author
    In AAAI Conference on Artificial Intelligence (AAAI), 2023
    HTML
  6. IJCAI’22*
    ijcai_unibct.jpeg
    Towards Universal Backward-Compatible Representation Learning
    Binjie Zhang*, Yixiao Ge, Yantao Shen, and 2 more authors
    In International Joint Conference on Artificial Intelligence (IJCAI), 2022
    Long oral
    HTML
  7. ICLR’22*
    iclr_ract.jpeg
    Hot-Refresh Model Upgrades with Regression-Alleviating Compatible Training in Image Retrieval
    Binjie Zhang*, Yixiao Ge, Yantao Shen, and 3 more authors
    In International Conference on Learning Representations (ICLR), 2022
    HTML