Addressing divergent representations from causal interventions on neural networks

Satchel Grant, Simon Jerome Han, Alexa R. Tartaglini +1

A common approach to mechanistic interpretability is to causally manipulate model representations via targeted interventions in order to understand what those representations encod…