Metraipolis › archive › dc_da8df4b00f79

Adversarial Self-Distillation — arXiv:2610.38732

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.38732
License (as archived)
CC0-1.0
Archived
2026-11-01T02:24:52.057Z
SHA-256 of the archived text
b1c0dcefa49094ac523292a4396173d69435b0e5c305a015747503fa1e84ee4a
Stable URL of this page
https://metraipolis.com/documento/dc_da8df4b00f79

How to cite

Adversarial Self-Distillation — arXiv:2610.38732 — arXiv (en), https://arxiv.org/abs/2610.38732 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_da8df4b00f79 (Metraipolis archive dc_da8df4b00f79, sha256 b1c0dcefa49094ac523292a4396173d69435b0e5c305a015747503fa1e84ee4a)
BibTeX
@misc{dcda8df4b00f79,
  title = {Adversarial Self-Distillation — arXiv:2610.38732},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.38732},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_da8df4b00f79; sha256 b1c0dcefa49094ac523292a4396173d69435b0e5c305a015747503fa1e84ee4a},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.38732. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Even though current safety-aligned language models are able to refuse harmful requests, their guard can still be bypassed by adversarial perturbations. As in other domains, adversarial training is considered a standard defense. However, due to the autoregressive nature of language generation, supervised finetuning on template refusals does not necessarily teach the model to refuse. Evidence of this is that the adversarially trained model still complies when the attacker injects a compliant prefill. We propose Adversarial Self-Distillation (ASD), which extends a model's own safe behavior to the space of adversarial perturbations. The teacher (the model itself given the clean prompt) can refuse the harmful request by default, or with the help of an additional instruction in its context. ASD distills this knowledge to the student (which is the same model under attack) at the prefixes that the student visits (due to on-policy sampling). These potentially include compliant openings. ASD achieves lower attack success than state-of-the-art defenses with little loss in multiple-choice accuracy, and it becomes able to recover a refusal even after a compliant opening.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.38732 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-01T02:12:44.132Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).