learn.ifs.name

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 “5 Conclusion and Future Work” · whole deck

5 Conclusion and Future Work

  1. What open question about defenses do Zou et al. raise?
    Whether adversarial training (attacking the model during fine-tuning and training on refusals) can make models robust without sacrificing capability, which has not been achieved for classical models, or whether pretraining itself needs to change.

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