Tag
This paper proposes a hybrid framework combining first-order safety alignment with zeroth-order refinement to enhance the robustness of LLM safety alignment against post-alignment perturbations. Theoretical and empirical results show that only a few refinement steps can improve robustness while preserving safety.
This paper systematically investigates unlearnable examples under diverse training paradigms, revealing that pretrained weights weaken existing methods, and proposes Shallow Semantic Camouflage (SSC) to maintain unlearnability by generating perturbations in a semantically valid subspace.