1
A Low-Rank Subspace Analysis of LLM Interventions
从副作用的根源剖析LLM行为干预，低秩子空间分析精准定位，为安全控制提供新视角
arXiv:2606.14388v1 Announce Type: new Abstract: Interventions designed to modify a particular behavior in LLMs, such as refusal or sycophancy, often p…