Metraipolis › archive › dc_7aea2044722e

Mood Matters: How Syntactic Sensitivity Undermines Safety Alignment — arXiv:2608.05409

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.05409
License (as archived)
CC0-1.0
Archived
2026-10-25T21:27:07.761Z
SHA-256 of the archived text
ebd133fc8e0028fa190ad109d300118638811f8d860f1ef55b5212bf0926bad9
Stable URL of this page
https://metraipolis.com/documento/dc_7aea2044722e

How to cite

Mood Matters: How Syntactic Sensitivity Undermines Safety Alignment — arXiv:2608.05409 — arXiv (en), https://arxiv.org/abs/2608.05409 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7aea2044722e (Metraipolis archive dc_7aea2044722e, sha256 ebd133fc8e0028fa190ad109d300118638811f8d860f1ef55b5212bf0926bad9)
BibTeX
@misc{dc7aea2044722e,
  title = {Mood Matters: How Syntactic Sensitivity Undermines Safety Alignment — arXiv:2608.05409},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.05409},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7aea2044722e; sha256 ebd133fc8e0028fa190ad109d300118638811f8d860f1ef55b5212bf0926bad9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.05409. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large language models typically undergo post-training to align them with safety policies but there exist many sophisticated jailbreaks that sidestep established safeguards. For instance, prior work by Andriushchenko et al. (2025) has found that changing the grammatical tense from present to past can be enough to elicit harmful responses. In this work, we uncover a more general failure of non-imperative syntactic forms. We demonstrate that this syntactic vulnerability exists in 16 models up to 70B parameters, using behavioral evaluation. To investigate the root cause, we apply causal mediation analysis, finding that refusal is partially conditioned on upstream syntactic features. By steering these purely syntactic features we are able to trigger and suppress refusal. Finally, we trace this ill-conditioning to linguistically biased post-training data of open-source models and show that increasing syntactic diversity can mitigate the issue. Our findings suggest that current alignment approaches introduce confounders that prevent a pure semantic grounding of the refusal decision.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-07 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.05409 — 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 2026-10-25T21:24:38.189Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).