Adaptive Probe-based Steering for Robust LLM Jailbreaking
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917512145272832 |
|---|---|
| author | Chen, Junxi Dong, Junhao Xie, Xiaohua |
| author_facet | Chen, Junxi Dong, Junhao Xie, Xiaohua |
| contents | Recent work has demonstrated the potential of contrastive steering for jailbreaking Large Language Models (LLMs). However, existing methods rely on limited and inherently biased contrastive prompts and require laborious manual tuning of steering strength, limiting their robustness and effectiveness. In this paper, we leverage the idea of model extraction to guide the learned steering vectors to approximate the ideal one and propose tuning the steering strength adaptively based on contrastive activations' statistics. Experiments demonstrate that our method notably improves the effectiveness and robustness of probe-based steering, without any extra contrastive prompts or laborious manual tuning. Being an attack paper, this paper focuses on revealing the breakdown of fortified LLMs, raising the average harmfulness score from 6\% to 70\%. Our code is available at https://github.com/fhdnskfbeuv/adaptiveSteering. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_20286 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Adaptive Probe-based Steering for Robust LLM Jailbreaking Chen, Junxi Dong, Junhao Xie, Xiaohua Cryptography and Security Machine Learning Recent work has demonstrated the potential of contrastive steering for jailbreaking Large Language Models (LLMs). However, existing methods rely on limited and inherently biased contrastive prompts and require laborious manual tuning of steering strength, limiting their robustness and effectiveness. In this paper, we leverage the idea of model extraction to guide the learned steering vectors to approximate the ideal one and propose tuning the steering strength adaptively based on contrastive activations' statistics. Experiments demonstrate that our method notably improves the effectiveness and robustness of probe-based steering, without any extra contrastive prompts or laborious manual tuning. Being an attack paper, this paper focuses on revealing the breakdown of fortified LLMs, raising the average harmfulness score from 6\% to 70\%. Our code is available at https://github.com/fhdnskfbeuv/adaptiveSteering. |
| title | Adaptive Probe-based Steering for Robust LLM Jailbreaking |
| topic | Cryptography and Security Machine Learning |
| url | https://arxiv.org/abs/2605.20286 |