Kejia Chen

Kejia Chen

Invisible
Nerd
Typing
Paradoxes

✨ About Me

I’m a Ph.D. candidate at the Visual Intelligence and Pattern Analysis (VIPA) Lab, Zhejiang University, supervised by Prof. Zunlei Feng and Prof. Mingli Song. Additionally, my research has been significantly shaped by the insightful support of Prof. Jian Lou and Prof. Ruoxi Jia.

Presently, I am a visiting student at Nanyang Technological University, working under the esteemed supervision of Prof. Tianwei Zhang.

πŸ”¬ Research Interests

  • Trustworthy Generative AI
    Exploring reliability, controllability, and ethical alignment in generative models, with a focus on safety and transparency.

  • Efficient Deep Learning
    Designing lightweight, high-performance neural architectures and training strategies to reduce computational cost without compromising accuracy.

I’ll be sharing research updates and progress here from time to timeβ€”stay tuned!

πŸ“ Publications

(* indicates equal contribution.)

  • PR 2026 Self-Improved Holistic Alignment for Preference Enhancement ｜Pdf ｜Code
    Kejia Chen, Jiawen Zhang, Jiazheng Yang, Zunlei Feng, Mingli Song
  • ICLR 2026 Safety at One Shot: Patching Fine-Tuned LLMs with A Single Instance ｜Pdf ｜Code
    Jiawen Zhang, Lipeng He, Kejia Chen, Jian Lou, Jian Liu, Xiaohu Yang, Ruoxi Jia
  • Usenix 2025 Activation Approximations Can Incur Safety Vulnerabilities in Aligned LLMs: Comprehensive Analysis and Defense ｜Pdf ｜Code
    Jiawen Zhang*, Kejia Chen*, Lipeng He, Jian Lou, Dan Li, Zunlei Feng, Mingli Song, Kui Ren, Xiaohu Yang
  • ICML 2025 Assessing Safety Risks and Quantization-aware Safety Patching for Quantized Large Language Models ｜Pdf ｜Code
    Kejia Chen*, Jiawen Zhang*, Jiacong Hu, Yu Wang, Mingli Song, Jian Lou, Zunlei Feng
  • ECAI 2024 SecPE: Secure Prompt Ensembling for Private and Robust Large Language Models ｜Pdf ｜Code
    Jiawen Zhang*, Kejia Chen*, Zunlei Feng, Mingli Song, Jian Liu, Jian Lou, Xiaohu Yang
  • NDSS 2025 Nexus: Secure Transformer Inference Made Non-interactive ｜Pdf ｜Code
    Jiawen Zhang, Xinpeng Yang, Lipeng He, Kejia Chen, Wen-jie Lu, Yinghao Wang, Xiaoyang Hou, Jian Liu, Kui Ren, Xiaohu Yang
  • NeurIPS 2024 Transformer-Doctor: Diagnosing and Treating Vision Transformers ｜Pdf ｜Code
    Jiacong Hu, Hao Chen, Kejia Chen, Yang Gao, Jingwen Ye, Xingen Wang, Mingli Song, Zunlei Feng
  • BCRA 2024 Private, Atomic, Incentive Mechanism for Federated Learning based on Blockchain ｜Pdf
    Kejia Chen*, Jiawen Zhang*, Xuanming Liu, Zunlei Feng, Xiaohu Yang

πŸ“– Review Services

I am honored to contribute to the research community as a technical reviewer for several premier conferences and journals in AI and Security.

My service includes:

  • Conference Program Committee: ICML, ICLR, NeurIPS, CVPR, ICCV, ECCV, ACL, EMNLP.
  • Journal Peer Review: IEEE Transactions on Image Processing (TIP), IEEE Transactions on Information Forensics and Security (TIFS).