Released AdverRobust, a open-source framework for adversarial training and robustness evaluation!
A PyTorch-based repository designed to provide a comprehensive framework for adversarial training and robustness evaluation in visual tasks.
🍟kejia Zhang
Master Student
Xiamen University
About Me
Welcome to my homepage! I am Kejia Zhang, a second-year Master’s student at Xiamen University’s Department of Artificial Intelligence. Prior to this, I received a Bachelor of Science in Mathematics from Anhui University in 2023. My research interests focus on Trustworthy Large Vision-Language Models (LVMs) and Adversarial Robustness.
I am conducting research in the Intelligent Multimedia Lab at Xiamen University, under the supervision of Professor Zhiming Luo.
Previously, I worked at the Institute of Intelligent Science and Network Security at Anhui University (2021-2023), under the supervision of Professor Haifeng Zhang.
Interests
- Artificial Intelligence
- Computer Vision
- Neural Network Robustness
Education
Department of Artificial Intelligence
Xiamen University
School of Mathematical Sciences
Anhui University
📚 My Research
My current research focuses on mitigating hallucinations in large vision-language models (LVMs), while leveraging my extensive experience in adversarial robustness, adversarial training, and adversarial attacks.
📩 (AD) I am seeking a PhD position for Fall 2026, with a focus on large model security and adversarial robustness. If you are aware of any opportunities that align with my research interests, please feel free to contact me. Thank you!
🚀 Featured Publications
🏖️ Publications
(2024).
Towards Adversarial Robustness via Debiased High-Confidence Logit Alignment.
Preprint.
(2024).
Mitigating Low-Frequency Bias: Feature Recalibration and Frequency Attention Regularization for Adversarial Robustness.
Preprint.
(2024).
A Collaborative Framework Using Multimodal Data and Adaptive Noise for Human Behavior Anomaly Detection.
International Joint Conference on Neural Networks (IJCNN).
(2024).
Harmonizing Feature Maps: A Graph Convolutional Approach for Enhancing Adversarial Robustness.
Preprint.
(2024).
Extracting Higher Order Topological Semantic via Motif-Based Deep Graph Neural Networks.
IEEE Transactions on Computational Social Systems (JCR-Q1).
Academic Meetings
Recent News