TY  - RPRT
TI  - TROJail: Trajectory-Level Optimization for Multi-Turn Large Language Model Jailbreaks with Process Rewards
AU  - Xiqiao Xiong
AU  - Ouxiang Li
AU  - Zhuo Liu
AU  - Moxin Li
AU  - Wentao Shi
AU  - Fengbin Zhu
AU  - Qifan Wang
AU  - Fuli Feng
PY  - 2026
UR  - https://arxiv.org/abs/2512.07761
ID  - 2512.07761
ER  -