TY  - RPRT
TI  - Short-length Adversarial Training Helps LLMs Defend Long-length Jailbreak Attacks: Theoretical and Empirical Evidence
AU  - Shaopeng Fu
AU  - Liang Ding
AU  - Jingfeng Zhang
AU  - Di Wang
PY  - 2026
UR  - https://arxiv.org/abs/2502.04204
ID  - 2502.04204
ER  -