Universal and Transferable Adversarial Attacks on Aligned Language Models
Zou et al., 2023, Universal and Transferable Adversarial Attacks on Aligned Language Models (arXiv preprint arXiv:2307.15043)
Subscriber deck — free during early access
1 cards in “3.1 Attacks on White-box Models” · whole deck
3.1 Attacks on White-box Models
- How effective was GCG on the open models it optimized against?Exact harmful strings: 88% on Vicuna-7B and 57% on LLaMA-2-7B-Chat (AutoPrompt 25% and 3%; PEZ and GBDA none). Harmful behaviors: 100% and 88%. A single suffix optimized on 25 behaviors also worked on held-out behaviors, so universal attacks exist.
- Zou et al., 2023, Universal and Transferable Adversarial Attacks on Aligned Language Models (arXiv preprint arXiv:2307.15043), §3.1
Source
- Andy Zou, Zifan Wang, Nicholas Carlini, Milad Nasr, J. Zico Kolter and Matt Fredrikson. 2023. Universal and Transferable Adversarial Attacks on Aligned Language Models. In arXiv preprint arXiv:2307.15043. https://arxiv.org/abs/2307.15043