TY  - RPRT
TI  - Defense-to-Attack: Bypassing Weak Defenses Enables Stronger Jailbreaks in Vision-Language Models
AU  - Yunhan Zhao
AU  - Xiang Zheng
AU  - Xingjun Ma
PY  - 2025
UR  - https://arxiv.org/abs/2509.12724
ID  - 2509.12724
ER  -