publications

2026

1. ICML’26

   

   REFLECTOR: Internalizing Step-wise Reflection against Indirect Jailbreak

   Jiachen Ma, Jiawen Zhang, Xiangtian Li, and 3 more authors

   ICML, 2026

   Abs PDF Code

   We propose Reflector, a framework that internalizes step-wise reflection to defend against sophisticated indirect jailbreak attacks.

2. Arxiv

   

   ReSaM: Representation-Level Safety Margin Alignment for Vision-Language Models

   Jiachen Ma, Jiawen Zhang, Chaochao Lu, and 1 more author

   Arxiv, 2026

   Abs PDF

   ReSaM achieves 94.6% safety with only 5 samples by aligning representation-level safety margins in low-rank subspaces.

2025

1. NAACL’25

   

   Jailbreaking Prompt Attack: A Controllable Adversarial Attack against Diffusion Models

   Jiachen Ma, Yijiang Li, Zhiqing Xiao, and 4 more authors

   In NAACL, 2025

   PDF

2023

1. ICCV’23

   

   CAME: Contrastive Automated Model Evaluation

   Ru Peng, Qiuyang Duan, Haobo Wang, and 5 more authors

   In ICCV, 2023

   PDF