FishBack: Pullback Fisher Geometry for Optimal Activation Steering in Transformers

Excerpt

arXiv:2605.17231v2 Announce Type: replace-cross Abstract: Activation steering has emerged as a lightweight approach for modifying language model behavior without parameter updates, yet existing methods remain brittle: unstable across layers and prone to disturbing behavior unrelated to the target concept. We trace these failures to a hidden assumption shared by widely-used methods such as CAA, ActAdd, and ITI: that the intermediate activation space is Euclidean. We show this assumption is fundam