notesum.ai

Published at November 4

Extracting Unlearned Information from LLMs with Activation Steering

cs.CL
cs.AI
cs.LG

Released Date: November 4, 2024

Authors: Atakan Seyitoğlu, Aleksei Kuvshinov1, Leo Schwinn1, Stephan Günnemann

Aff.: 1Department of Computer Science & Munich Data Science Institute, Technical University of Munich

Arxiv: http://arxiv.org/abs/2411.02631v1