Metraipolis › archive › dc_50b995cb301b
Why Jailbreaks Succeed in Diffusion Language Models: An Energy Landscape Analysis — arXiv:2609.30841
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.30841
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T01:38:22.798Z
- SHA-256 of the archived text
c9cbcd8bec4a323e091ce8f334e87c1c3e9b0ebd3896f788eb35fa8c5a6c14d9- Stable URL of this page
https://metraipolis.com/documento/dc_50b995cb301b
How to cite
Why Jailbreaks Succeed in Diffusion Language Models: An Energy Landscape Analysis — arXiv:2609.30841 — arXiv (en), https://arxiv.org/abs/2609.30841 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_50b995cb301b (Metraipolis archive dc_50b995cb301b, sha256 c9cbcd8bec4a323e091ce8f334e87c1c3e9b0ebd3896f788eb35fa8c5a6c14d9)
BibTeX
@misc{dc50b995cb301b,
title = {Why Jailbreaks Succeed in Diffusion Language Models: An Energy Landscape Analysis — arXiv:2609.30841},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.30841},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_50b995cb301b; sha256 c9cbcd8bec4a323e091ce8f334e87c1c3e9b0ebd3896f788eb35fa8c5a6c14d9},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.30841. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Existing attacks and defenses for diffusion-based large language models (dLLMs) target specific vulnerabilities but lack a shared framework explaining why attacks succeed. We propose one by interpreting safety alignment as shaping the denoising energy landscape: a well-aligned model routes harmful queries toward safe outputs through an energy barrier that separates the two regions. Current jailbreak attacks reduce to two strategies for circumventing this barrier: obscuring the query's safety disposition at initialisation, or intervening mid-trajectory to force the denoising path across the energy barrier. From this perspective and the result that masked diffusion models minimise kinetic energy during denoising, we derive three complementary, training-free detection signals: a step-0 ratio that reads the initial safety disposition from the logit distribution before generation begins, and two trajectory-velocity signals that track kinetic energy in complementary subspaces of the logit space. An attack must either reveal its intent at initialisation or expend kinetic energy to cross the barrier in at least one monitored subspace, so the three signals cover each other's blind spots in the energy budget by construction. Evaluation across three dense dLLMs (LLaDA-8B, LLaDA-1.5, Dream-7B) and a sparse mixture-of-experts dLLM (LLaDA-MoE-7B) confirms this complementarity. In stress tests of known attacks, every configuration that evades detection also fails to produce harmful content, suggesting that the detection and barrier-crossing thresholds are hard to separate. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.30841 — 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 2026-10-08T01:12:14.297Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).