Improving Alignment and Robustness with Circuit Breakers
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866929418767695872 |
|---|---|
| author | Zou, Andy Phan, Long Wang, Justin Duenas, Derek Lin, Maxwell Andriushchenko, Maksym Wang, Rowan Kolter, Zico Fredrikson, Matt Hendrycks, Dan |
| author_facet | Zou, Andy Phan, Long Wang, Justin Duenas, Derek Lin, Maxwell Andriushchenko, Maksym Wang, Rowan Kolter, Zico Fredrikson, Matt Hendrycks, Dan |
| contents | AI systems can take harmful actions and are highly vulnerable to adversarial attacks. We present an approach, inspired by recent advances in representation engineering, that interrupts the models as they respond with harmful outputs with "circuit breakers." Existing techniques aimed at improving alignment, such as refusal training, are often bypassed. Techniques such as adversarial training try to plug these holes by countering specific attacks. As an alternative to refusal training and adversarial training, circuit-breaking directly controls the representations that are responsible for harmful outputs in the first place. Our technique can be applied to both text-only and multimodal language models to prevent the generation of harmful outputs without sacrificing utility -- even in the presence of powerful unseen attacks. Notably, while adversarial robustness in standalone image recognition remains an open challenge, circuit breakers allow the larger multimodal system to reliably withstand image "hijacks" that aim to produce harmful content. Finally, we extend our approach to AI agents, demonstrating considerable reductions in the rate of harmful actions when they are under attack. Our approach represents a significant step forward in the development of reliable safeguards to harmful behavior and adversarial attacks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_04313 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Improving Alignment and Robustness with Circuit Breakers Zou, Andy Phan, Long Wang, Justin Duenas, Derek Lin, Maxwell Andriushchenko, Maksym Wang, Rowan Kolter, Zico Fredrikson, Matt Hendrycks, Dan Machine Learning Artificial Intelligence Computation and Language Computer Vision and Pattern Recognition Computers and Society AI systems can take harmful actions and are highly vulnerable to adversarial attacks. We present an approach, inspired by recent advances in representation engineering, that interrupts the models as they respond with harmful outputs with "circuit breakers." Existing techniques aimed at improving alignment, such as refusal training, are often bypassed. Techniques such as adversarial training try to plug these holes by countering specific attacks. As an alternative to refusal training and adversarial training, circuit-breaking directly controls the representations that are responsible for harmful outputs in the first place. Our technique can be applied to both text-only and multimodal language models to prevent the generation of harmful outputs without sacrificing utility -- even in the presence of powerful unseen attacks. Notably, while adversarial robustness in standalone image recognition remains an open challenge, circuit breakers allow the larger multimodal system to reliably withstand image "hijacks" that aim to produce harmful content. Finally, we extend our approach to AI agents, demonstrating considerable reductions in the rate of harmful actions when they are under attack. Our approach represents a significant step forward in the development of reliable safeguards to harmful behavior and adversarial attacks. |
| title | Improving Alignment and Robustness with Circuit Breakers |
| topic | Machine Learning Artificial Intelligence Computation and Language Computer Vision and Pattern Recognition Computers and Society |
| url | https://arxiv.org/abs/2406.04313 |