Saved in:
Bibliographic Details
Main Authors: Zaremba, Wojciech, Nitishinskaya, Evgenia, Barak, Boaz, Lin, Stephanie, Toyer, Sam, Yu, Yaodong, Dias, Rachel, Wallace, Eric, Xiao, Kai, Heidecke, Johannes, Glaese, Amelia
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2501.18841
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909470970347520
author Zaremba, Wojciech
Nitishinskaya, Evgenia
Barak, Boaz
Lin, Stephanie
Toyer, Sam
Yu, Yaodong
Dias, Rachel
Wallace, Eric
Xiao, Kai
Heidecke, Johannes
Glaese, Amelia
author_facet Zaremba, Wojciech
Nitishinskaya, Evgenia
Barak, Boaz
Lin, Stephanie
Toyer, Sam
Yu, Yaodong
Dias, Rachel
Wallace, Eric
Xiao, Kai
Heidecke, Johannes
Glaese, Amelia
contents We conduct experiments on the impact of increasing inference-time compute in reasoning models (specifically OpenAI o1-preview and o1-mini) on their robustness to adversarial attacks. We find that across a variety of attacks, increased inference-time compute leads to improved robustness. In many cases (with important exceptions), the fraction of model samples where the attack succeeds tends to zero as the amount of test-time compute grows. We perform no adversarial training for the tasks we study, and we increase inference-time compute by simply allowing the models to spend more compute on reasoning, independently of the form of attack. Our results suggest that inference-time compute has the potential to improve adversarial robustness for Large Language Models. We also explore new attacks directed at reasoning models, as well as settings where inference-time compute does not improve reliability, and speculate on the reasons for these as well as ways to address them.
format Preprint
id arxiv_https___arxiv_org_abs_2501_18841
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Trading Inference-Time Compute for Adversarial Robustness
Zaremba, Wojciech
Nitishinskaya, Evgenia
Barak, Boaz
Lin, Stephanie
Toyer, Sam
Yu, Yaodong
Dias, Rachel
Wallace, Eric
Xiao, Kai
Heidecke, Johannes
Glaese, Amelia
Machine Learning
Cryptography and Security
We conduct experiments on the impact of increasing inference-time compute in reasoning models (specifically OpenAI o1-preview and o1-mini) on their robustness to adversarial attacks. We find that across a variety of attacks, increased inference-time compute leads to improved robustness. In many cases (with important exceptions), the fraction of model samples where the attack succeeds tends to zero as the amount of test-time compute grows. We perform no adversarial training for the tasks we study, and we increase inference-time compute by simply allowing the models to spend more compute on reasoning, independently of the form of attack. Our results suggest that inference-time compute has the potential to improve adversarial robustness for Large Language Models. We also explore new attacks directed at reasoning models, as well as settings where inference-time compute does not improve reliability, and speculate on the reasons for these as well as ways to address them.
title Trading Inference-Time Compute for Adversarial Robustness
topic Machine Learning
Cryptography and Security
url https://arxiv.org/abs/2501.18841