Extracting Unlearned Information from LLMs with Activation Steering

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Seyitoğlu, Atakan, Kuvshinov, Aleksei, Schwinn, Leo, Günnemann, Stephan
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929576510226432
author Seyitoğlu, Atakan
Kuvshinov, Aleksei
Schwinn, Leo
Günnemann, Stephan
author_facet Seyitoğlu, Atakan
Kuvshinov, Aleksei
Schwinn, Leo
Günnemann, Stephan
contents An unintended consequence of the vast pretraining of Large Language Models (LLMs) is the verbatim memorization of fragments of their training data, which may contain sensitive or copyrighted information. In recent years, unlearning has emerged as a solution to effectively remove sensitive knowledge from models after training. Yet, recent work has shown that supposedly deleted information can still be extracted by malicious actors through various attacks. Still, current attacks retrieve sets of possible candidate generations and are unable to pinpoint the output that contains the actual target information. We propose activation steering as a method for exact information retrieval from unlearned LLMs. We introduce a novel approach to generating steering vectors, named Anonymized Activation Steering. Additionally, we develop a simple word frequency method to pinpoint the correct answer among a set of candidates when retrieving unlearned information. Our evaluation across multiple unlearning techniques and datasets demonstrates that activation steering successfully recovers general knowledge (e.g., widely known fictional characters) while revealing limitations in retrieving specific information (e.g., details about non-public individuals). Overall, our results demonstrate that exact information retrieval from unlearned models is possible, highlighting a severe vulnerability of current unlearning techniques.
format Preprint
id arxiv_https___arxiv_org_abs_2411_02631
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Extracting Unlearned Information from LLMs with Activation Steering
Seyitoğlu, Atakan
Kuvshinov, Aleksei
Schwinn, Leo
Günnemann, Stephan
Computation and Language
Artificial Intelligence
Machine Learning
An unintended consequence of the vast pretraining of Large Language Models (LLMs) is the verbatim memorization of fragments of their training data, which may contain sensitive or copyrighted information. In recent years, unlearning has emerged as a solution to effectively remove sensitive knowledge from models after training. Yet, recent work has shown that supposedly deleted information can still be extracted by malicious actors through various attacks. Still, current attacks retrieve sets of possible candidate generations and are unable to pinpoint the output that contains the actual target information. We propose activation steering as a method for exact information retrieval from unlearned LLMs. We introduce a novel approach to generating steering vectors, named Anonymized Activation Steering. Additionally, we develop a simple word frequency method to pinpoint the correct answer among a set of candidates when retrieving unlearned information. Our evaluation across multiple unlearning techniques and datasets demonstrates that activation steering successfully recovers general knowledge (e.g., widely known fictional characters) while revealing limitations in retrieving specific information (e.g., details about non-public individuals). Overall, our results demonstrate that exact information retrieval from unlearned models is possible, highlighting a severe vulnerability of current unlearning techniques.
title Extracting Unlearned Information from LLMs with Activation Steering
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2411.02631