# Revisiting Jailbreaking for Large Language Models: A Representation Engineering Perspective

Source: [Revisiting Jailbreaking for Large Language Models: A Representation Engineering Perspective](<https://www.sota2.com/research/paper/68fa770b1e209131cea70b01>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T04:08:58.245Z

Authors: Tianlong Li, Zhenghua Wang, Wenhao Liu, Muling Wu, Shihan Dou, Changze Lv, Xiaohua Wang, Xiaoqing Zheng, Xuanjing Huang

Published: 2024-01-12T00:50:04.000Z

## Abstract

The recent surge in jailbreaking attacks has revealed significant vulnerabilities in Large Language Models (LLMs) when exposed to malicious inputs. While various defense strategies have been proposed to mitigate these threats, there has been limited research into the underlying mechanisms that make LLMs vulnerable to such attacks. In this study, we suggest that the self-safeguarding capability of LLMs is linked to specific activity patterns within their representation space. Although these patterns have little impact on the semantic content of the generated text, they play a crucial role in shaping LLM behavior under jailbreaking attacks. Our findings demonstrate that these patterns can be detected with just a few pairs of contrastive queries. Extensive experimentation shows that the robustness of LLMs against jailbreaking can be manipulated by weakening or strengthening these patterns. Further visual analysis provides additional evidence for our conclusions, providing new insights into the jailbreaking phenomenon. These findings highlight the importance of addressing the potential misuse of open-source LLMs within the community.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Jailbreak Attack on Advbench Harmful Behaviors (subset 100 samples)](<https://www.sota2.com/research/sota/jailbreak-attack-on-advbench-harmful-behaviors-subset-100-samples.md>): Jailbreak Attack; Advbench Harmful Behaviors (subset 100 samples); Attack Success Rate AK: 83.54.

- [Jailbreak Attack on StrongREJECT 2024 (Full)](<https://www.sota2.com/research/sota/jailbreak-attack-on-strongreject-2024-full.md>): Jailbreak Attack; StrongREJECT 2024 (Full); AK: 84.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2401.06824>)

[PDF](<https://arxiv.org/pdf/2401.06824.pdf>)