Skip to main content
Understanding Unreliability of Steering Vectors in Language Models: Geometric Predictors and the Limits of Linear Approximations | Signal Canvas | ScienceToStartup