Robust LLM safeguarding via refusal feature adversarial training
Keywords: large language models, adversarial robustness, representation analysis, mechanistic interpretability
TL;DR: Mechanistic analysis and mitigation of LLM vulnerability against adversarial attacks
Abstract: Large language models (LLMs) are vulnerable to adversarial attacks that can elicit harmful responses. Defending against such attacks remains challenging due to the opacity of jailbreaking mechanisms and the high computational cost of training LLMs robustly. We demonstrate that adversarial attacks share a universal mechanism for circumventing LLM safeguards that works by ablating a dimension in the residual stream embedding space called the refusal feature. We further show that the operation of refusal feature ablation (RFA) approximates the worst-case perturbation of offsetting model safety. Based on these findings, we propose Refusal Feature Adversarial Training (ReFAT), a novel algorithm that efficiently performs LLM adversarial training by simulating the effect of input-level attacks via RFA. Experiment results show that ReFAT significantly improves the robustness of three popular LLMs against a wide range of adversarial attacks, with considerably less computational overhead compared to existing adversarial training methods.
Primary Area: interpretability and explainable AI
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 6962
Loading