Steering Along Manifolds to Control Neural Networks
Wurgaft, Goodman, Rager, Fel, Kowal, Geiger, Shyam, Lubana, Feucht, Bhalla, Haklay, Bigelow, Sarfati, McGrath, Lewis, Merullo · 2025
Steering along curved representation geometries gives more effective behavioural control than standard linear steering.
mech-interpinterpretabilitymanifoldssteering