Chengyi CAI (蔡澄奕)

Ph.D. Student

School of Computing and Information Systems,     Faculty of Engineering and Information Technology, The University of Melbourne

Address: Melbourne Connect, 700 Swanston Street, Carlton, VIC, Australia

GitHub Profile      Google Scholar Profile

...

About Me

Hi there, I am Chengyi Cai, a second-year Ph.D. student at Trustworthy Machine Learning and Reasoning Group (TMLR) in the Faculty of Engineering and Information Technology, the University of Melbourne, luckily advised by Dr. Feng Liu, A/Prof Jianzhong Qi and Dr. Zesheng Ye. I was once supervised by Dr. Yuchen Guo and Prof. Qionghai Dai, and obtained a master's degree from Tsinghua University. My current research interests lie in Learning/Inference under Distribution Shift, including:

  • Model reprogramming of pre-trained vision models and VLMs (Vision-Language Models)
  • Test-time alignment of VLMs and MLLMs (Multimodal Large Language Models)

    • Recent News

    Selected Publications

    Image
    ICLR 2025

    Attribute-based Visual Reprogramming for Vision-Language Models. [Code]

    Chengyi Cai, Zesheng Ye, Lei Feng, Jianzhong Qi, Feng Liu*. [ICLR 2025]

    Image
    NeurIPS 2024

    Bayesian-guided Label Mapping for Visual Reprogramming. [Code]

    Chengyi Cai, Zesheng Ye, Lei Feng, Jianzhong Qi, Feng Liu*. [NeurIPS 2024 Oral]

    Image
    ICML 2024

    Sample-specific Masks for Visual Reprogramming-based Prompting. [Code]

    Chengyi Cai, Zesheng Ye, Lei Feng, Jianzhong Qi, Feng Liu*. [ICML 2024 Spotlight]

    Academic Services

    Research Intern

    Awards & Competitions