Metraipolis › archive › dc_c51ec957505f
ECHO: Replacing the Compliance Proxy with Response Harmfulness in Optimization-Based Jailbreak Attack — arXiv:2610.23030
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.23030
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T08:44:05.563Z
- SHA-256 of the archived text
e1b3157c237ee90a969dbccff350700995a09685f3cfe2a72d511bc38ad41507- Stable URL of this page
https://metraipolis.com/documento/dc_c51ec957505f
How to cite
ECHO: Replacing the Compliance Proxy with Response Harmfulness in Optimization-Based Jailbreak Attack — arXiv:2610.23030 — arXiv (en), https://arxiv.org/abs/2610.23030 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c51ec957505f (Metraipolis archive dc_c51ec957505f, sha256 e1b3157c237ee90a969dbccff350700995a09685f3cfe2a72d511bc38ad41507)
BibTeX
@misc{dcc51ec957505f,
title = {ECHO: Replacing the Compliance Proxy with Response Harmfulness in Optimization-Based Jailbreak Attack — arXiv:2610.23030},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.23030},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c51ec957505f; sha256 e1b3157c237ee90a969dbccff350700995a09685f3cfe2a72d511bc38ad41507},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.23030. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Despite extensive safety alignment, large language models remain vulnerable to jailbreak attacks. Existing optimization-based attacks share a single objective: maximize a token-level proxy of victim compliance---such as the likelihood of an affirmative prefix---computed on the prompt rather than the decoded reply. This proxy cannot distinguish a cautious hedge from a fully actionable harmful answer, leaving the search blind to what the model actually says. We propose ECHO (Evaluation-Conditioned Harmfulness Optimization), which replaces the compliance proxy with response-level, continuous harmfulness scores from HarmSig, a 1.7B regressor grounded in actual harmful capability. ECHO couples HarmSig with five word-level perturbation operators and Keyword-Anchored Beam Search with global top-m memory, under a strict black-box threat model. Across five open-weight victims on AdvBench, ECHO achieves up to 0.898 StrongREJECT mean score and 89.23% HarmBench success. Our code is publicly available at this https URL . Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Cryptography and Security (cs.CR) ; Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2610.23030 — 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 2026-10-19T08:39:24.901Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).