|
I am a PhD candidate at the Department of Electrical and Computer Engineering at UC, Riverside, where I work on trustworthy machine learning and inverse problems. I am fortunate to be advised by Professor M. Salman Asif. Prior to joining UCR, I received my Bachelor's degree from Huazhong University of Science and Technology in 2022.
CV /
Scholar /
GitHub /
Hugging Face /
Lab |
|
|
Jul. 2025: I will present our work SLUG at ICML 2025 (Vancouver, BC). |
|
Zikui Cai, Yaoteng Tan, M. Salman Asif ICML, 2025 (New) arXiv / code / project page We propose a highly efficient machine unlearning method for fundation models (e.g., CLIP, Stable Diffusion, VLMs) that requires only one-time gradient calculation and one-step update on one model layer that are selected based on introduced metrics, layer importance and gradient alignment. Our method provide a modular modification for post-training large models. |
|
Zikui Cai*, Yaoteng Tan*, M. Salman Asif (* Equal contribution) CVPR, 2023 arXiv / open access / code / poster We propose a query-efficient approach for adversarial attacks on dense prediction models. Our proposed method can generate a single perturbation that can fool multiple blackbox detection and segmentation models simultaneously, demonstrate generalizability across different tasks. |
|
|
|
Yaoteng Tan, Zikui Cai, M. Salman Asif Preprint, 2024 (New) arXiv Many properties of adversarial attacks are well-studied today (e.g., optimization, transferability, physical implementa- tion, etc.). In this work, we explore an under-researched transform-dependent property of adversarial attacks, which the optimization process of additive adversarial perturbations can be combined with various image transformations to produce versatile, transform-dependent attack effects. |
|
Conference review:
Teaching Assistant:
Acknowledgement: template from Jon Barron |
