Seonglae ChoRedefining inner product so that causally separated (independent) features are orthogonal to each other
The Linear Representation Hypothesis and the Geometry of Large...
Informally, the 'linear representation hypothesis' is the idea that high-level concepts are represented linearly as directions in some representation space. In this paper, we address two closely...
https://openreview.net/forum?id=T0PoOJg8cK¬eId=hIzVPo3wws