Tianrui Zhu (朱添瑞)

I am a master's student at Tsinghua University, advised by Prof. Yansong Tang. I received my B.S. in Communication Engineering from Wuhan University in 2025.

I aim to advance AI systems that can learn from experience and act effectively in both digital and physical environments. My current focus is on self-improving agents: how models can identify their own limitations, generate useful training data, and turn experience into stronger capabilities.

Email  /  Github

profile photo

Education

dise Tsinghua University, 2025.09 - Now
  • M.S. degree in Data Science
  • Shenzhen International Graduate School
    • dise Wuhan University, 2021.09 - 2025.06
  • B.S. degree in Communication Engineering
  • Electronic Information School

    • Industrial Experience

    JD.com logo JD.com, Research Intern. 2026.06 - Present
  • Evaluation and inference infrastructure for multimodal foundation models.
  • Synthetic data generation for post-training.
  • Agentic online reinforcement learning.
    • dise ChinaTelecom TeleAI, Shanghai, China. 2024.10 - 2025.02
  • Project: Efficient inference framework of LLMs.
  • Work with Dr. JiaWei Shao
    • News

  • 2025-06: 1 paper about Image Editing was accepted on ICCV 2025 .

    • Recent Selected Publications

    (*Equal Contribution, #Corresponding Author)
    dise KV-Edit: Training-Free Image Editing for Precise Background Preservation
    Tianrui Zhu*, Shiyi Zhang*, Jiawei Shao, Yansong Tang#
    IEEE International Conference on Computer Vision (ICCV), 2025

    [arXiv] [Code]

    We propose KV-Edit to address the challenge of background preservation in image editing, thereby enhancing the practicality of AI editing.

    dise Memorize-and-Generate: Towards Long-Term Consistency in Real-Time Video Generation
    Tianrui Zhu*, Shiyi Zhang*, Zhirui Sun, Jingqi Tian, Yansong Tang#
    arXiv preprint, 2025

    [arXiv] [Code]

    We explore MAG framework, which decouples memory and generation into independent tasks, aiming to enhance historical scene consistency while achieving efficient compression for GPU memory.

    Selected Honors and Awards

  • National Schoolship, 2024.

  • The third prize - National Undergraduate Electronics Design Contest(Analog Electronic System), 2024.

  • The second prize - National Undergraduate Electronics Design Contest, 2023.