notesum.ai

Published at November 4

Improving Steering Vectors by Targeting Sparse Autoencoder Features

cs.LG
cs.AI
cs.CL

Released Date: November 4, 2024

Authors: Sviatoslav Chalnev1, Matthew Siu1, Arthur Conmy1

Aff.: 1Not specified

Arxiv: http://arxiv.org/abs/2411.02193v1