The Linear Representation Hypothesis and the Geometry of Large Language Models

Open in new window