Zikui Cai

I am a postdoctoral associate at University of Maryland, College Park working with Furong Huang and Tom Goldstein. I obtained my Ph.D. from Electrical and Computer Engineering at the University of California, Riverside, under the guidance of M. Salman Asif. Originally from Wuhan, the "River City" of China, I completed my B.S. in Automation at Wuhan University of Technology. I am currently on the job market for faculty positions starting in Fall 2026.

My research focuses on trustworthy machine learning, where I strive to identify vulnerabilities and enhance the trustworthiness of machine learning systems. My long-term vision is to unify robust perception, executable reasoning, and safe embodiment into a coherent framework for autonomous agents. I aim to build systems that do not merely predict the next token, but actively reconstruct their environment, verify their plans through mental or physical simulation, and self-improve safely in the wild.

Research Interests

  • Multi-modal reasoning across vision, language, and action [ MORSE-500 ]
  • Building AI agents that plan, act, and self-improve
  • Embodied AI for navigation and manipulation in robotics
  • Agentic safety and alignment in multi-agent ecosystems [ AegisLLM ]
  • Controllable and safe AI via model editing and unlearning [ SLUG, Robust Eval ]
  • Exposing model vulnerabilities via efficient blackbox attacks [ BASES, EBAD, Context-Aware-Attacks, ZQA ]
  • Detecting adversarial attacks using language models [ SCENE-Lang ]
  • Enhancing user privacy via de-identification while preserving utility [ Disguise ]
  • Improving physical modeling via end-to-end optimization [ Learn PR, Learn CDI ]

News

Selected Publications

Targeted Unlearning with Single Layer Unlearning Gradient
Zikui Cai, Yaoteng Tan, M Salman Asif
ICML 2025
arxiv code
AegisLLM: Scaling Agentic Systems for Self-Reflective Defense in LLM Security
Zikui Cai, Shayan Shabihi, Bang An, Zora Che, Brian R. Bartoldson, Bhavya Kailkhura, Tom Goldstein, Furong Huang
ICLR 2025 BuildingTrust
arxiv code
Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities
Zora Che, Stephen Casper, Robert Kirk, et al.
TMLR 2025
arxiv code
Can Textual Unlearning Solve Cross-Modality Safety Alignment?
Trishna Chakraborty, Erfan Shayegani, Zikui Cai, et al.
EMNLP 2024
arxiv
Disguise without Disruption: Utility-Preserving Face De-Identification
Zikui Cai, Zhongpai Gao, Benjamin Planche, Meng Zheng, Terrence Chen, M. Salman Asif, Ziyan Wu
AAAI 2024
arxiv
Ensemble-based Blackbox Attacks on Dense Prediction
Zikui Cai*, Yaoteng Tan*, M. Salman Asif
CVPR 2023
arxiv code
Blackbox Attacks via Surrogate Ensemble Search
Zikui Cai, Chengyu Song, Srikanth V. Krishnamurthy, Amit K. Roy-Chowdhury, M. Salman Asif
NeurIPS 2022
arxiv code poster slides talk
Zero-Query Transfer Attacks on Context-Aware Object Detectors
Zikui Cai, Shantanu Rane, Alejandro E. Brito, et al.
CVPR 2022
arxiv open access
Context-Aware Transfer Attacks for Object Detection
Zikui Cai, Xinxin Xie, Shasha Li, et al.
AAAI 2022
arxiv code slides poster talk
Exploiting Multi-Object Relationships for Detecting Adversarial Attacks in Complex Scenes
Mingjun Yin, Shasha Li, Zikui Cai, et al.
ICCV 2021
arxiv open access
Data-Driven Illumination Patterns For Coded Diffraction Imaging
Zikui Cai, Rakib Hyder, M. Salman Asif
ICIP 2021
paper
Solving Phase Retrieval with a Learned Reference
Rakib Hyder*, Zikui Cai*, M. Salman Asif
ECCV 2020
arxiv code