2025 "robust generalization" Papers

Short-length Adversarial Training Helps LLMs Defend Long-length Jailbreak Attacks: Theoretical and Empirical Evidence

Shaopeng Fu, Liang Ding, Jingfeng ZHANG et al.