1
The Geometry of Refusal: Linear Instability in Safety-Aligned LLMs
从几何角度揭示安全对齐LLM的拒绝机制实为线性特征，提出对比对数引导方法验证其不稳定性。
arXiv:2606.22686v1 Announce Type: cross Abstract: Modern Large Language Models (LLMs) rely on extensive safety alignment, yet the mechanistic basis of…