Metraipolis › archive › dc_719fbb859bc0

Reactivating Alignment: Defending LLMs from Jailbreaks via Intention-Aware Input-Output Matching — arXiv:2610.04470

Source
arXiv
Original
https://arxiv.org/abs/2610.04470
License (as archived)
CC0-1.0
Archived
2026-10-06T05:24:34.931Z
SHA-256 of the archived text
180b0e31cd8663fc5851042cd2f095610a09dfed7cebb90454dc7c0ec51a2f54
Stable URL of this page
https://metraipolis.com/documento/dc_719fbb859bc0

How to cite

Reactivating Alignment: Defending LLMs from Jailbreaks via Intention-Aware Input-Output Matching — arXiv:2610.04470 — arXiv, https://arxiv.org/abs/2610.04470 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_719fbb859bc0 (Metraipolis archive dc_719fbb859bc0, sha256 180b0e31cd8663fc5851042cd2f095610a09dfed7cebb90454dc7c0ec51a2f54)
BibTeX
@misc{dc719fbb859bc0,
  title = {Reactivating Alignment: Defending LLMs from Jailbreaks via Intention-Aware Input-Output Matching — arXiv:2610.04470},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.04470},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_719fbb859bc0; sha256 180b0e31cd8663fc5851042cd2f095610a09dfed7cebb90454dc7c0ec51a2f54},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.04470. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large language models (LLMs) remain vulnerable to jailbreak attacks that conceal harmful intent within complex adversarial prompts. Existing defenses primarily rely on input perturbation or harmful-output suppression, but they rarely model where malicious intent resides, resulting in brittle protection and excessive over-refusal. We propose SENTINEL, a plug-and-play, generation-time jailbreak defense that reframes mitigation as an intent extraction problem. Our key insight is that instruction-tuned LLMs exhibit strong input--output semantic consistency: regardless of jailbreak complexity, generated outputs tend to align with the attacker's true intent. SENTINEL exploits this property by matching semantically aligned input--output regions to extract intention-revealing subsequences, scores these subsequences using refusal-direction projections to estimate harmfulness, and halts generation when necessary. Experiments on HarmBench across multiple LLMs show that SENTINEL reduces jailbreak success rates to close to 5\% while maintaining low over-refusal. We further demonstrate robustness to adaptive attacks and provide a mechanistic interpretation: SENTINEL re-distributes jailbreak features from alignment blind spots to aligned regions.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2610.04470 — 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 2026-10-06T05:24:05.476Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).