Metraipolis › archive › dc_f4a00ea277ee
J-Anchor: Defending Against Prefill Attacks via J-Space Selective Refusal Restoration — arXiv:2611.01342
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.01342
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T18:17:51.187Z
- SHA-256 of the archived text
324d315c87b143ec1d491c07aa49c63362b6aab3e492747f70e16d2d0aaf38d5- Stable URL of this page
https://metraipolis.com/documento/dc_f4a00ea277ee
How to cite
J-Anchor: Defending Against Prefill Attacks via J-Space Selective Refusal Restoration — arXiv:2611.01342 — arXiv (en), https://arxiv.org/abs/2611.01342 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f4a00ea277ee (Metraipolis archive dc_f4a00ea277ee, sha256 324d315c87b143ec1d491c07aa49c63362b6aab3e492747f70e16d2d0aaf38d5)
BibTeX
@misc{dcf4a00ea277ee,
title = {J-Anchor: Defending Against Prefill Attacks via J-Space Selective Refusal Restoration — arXiv:2611.01342},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.01342},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f4a00ea277ee; sha256 324d315c87b143ec1d491c07aa49c63362b6aab3e492747f70e16d2d0aaf38d5},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.01342. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Prefill jailbreaks can cause safety-aligned language models to continue an unsafe assistant prefix even when they would refuse the same request under normal decoding. We study how the model's internal harmfulness and refusal signals evolve along this attacker-controlled trajectory. Using the Jacobian lens, we track vocabulary-grounded harmfulness and refusal signals from the assistant header through prefilling and subsequent generation. We identify a characteristic phenomenon, which we call refusal-state erosion: both the refusal signal and the harmfulness signal weaken as the prefill progresses, but the refusal signal progressively declines to the benign baseline by the end of the prefill, while the harmfulness signal remains above its benign baseline and detectable during subsequent generation. Motivated by this observation, we introduce J-Anchor, a training-free inference-time defense that uses the model's pre-attack harmfulness and refusal signals as a request-specific anchor and selectively restores weakened refusal in J-space during prefilling and decoding. J-Anchor substantially reduces attack success under both fixed and deep prefill attacks across different model families. J-Anchor largely preserves benign behavior and reasoning ability, and ablations highlight the importance of targeted restoration in the J-space refusal subspace. Our results suggest that preserving latent refusal dynamics offers a promising route to robust defense against prefill jailbreaks. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2611.01342 — 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 2026-11-03T18:09:53.582Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).