Metraipolis › archive › dc_0130da192f0a

A Safe Prototype Is Not a Safety Direction: Reference Dependence and Prompt Confounds in Response-Safety Embeddings — arXiv:2610.01801

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.01801
License (as archived)
CC0-1.0
Archived
2026-10-06T13:57:28.570Z
SHA-256 of the archived text
b1eb019df4f27d0db67862db9573166316034f6ac2b72e0dc8699b063e7aebf4
Stable URL of this page
https://metraipolis.com/documento/dc_0130da192f0a

How to cite

A Safe Prototype Is Not a Safety Direction: Reference Dependence and Prompt Confounds in Response-Safety Embeddings — arXiv:2610.01801 — arXiv (en), https://arxiv.org/abs/2610.01801 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0130da192f0a (Metraipolis archive dc_0130da192f0a, sha256 b1eb019df4f27d0db67862db9573166316034f6ac2b72e0dc8699b063e7aebf4)
BibTeX
@misc{dc0130da192f0a,
  title = {A Safe Prototype Is Not a Safety Direction: Reference Dependence and Prompt Confounds in Response-Safety Embeddings — arXiv:2610.01801},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.01801},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0130da192f0a; sha256 b1eb019df4f27d0db67862db9573166316034f6ac2b72e0dc8699b063e7aebf4},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.01801. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Can response safety be scored by cosine similarity to the mean embedding of known-safe responses? A recent sleeper-agent detector proposes exactly this score, yet the raw positive-centroid rule is not identified: positive observations locate the safe class relative to an encoder origin, but do not determine which direction separates safe from unsafe responses. We audit the rule on two prompt-controlled, human-labeled corpora and one auxiliary jury-labeled source control, using four frozen encoders and prompt-grouped splits. On the human-labeled corpora the safe prototype reaches ROC-AUC 0.457-0.545, with two cells significantly below chance and one above, while an explicit safe-minus-unsafe reference reaches 0.588-0.738 on the same embeddings; on the jury control the prototype is inverted (0.358-0.405) and the reference reaches 0.754-0.793. At validation-calibrated 5% false-safe thresholds, the reference accepts more safe responses on PKU-SafeRLHF (0.153-0.263 versus 0.039-0.061 across encoders) and Aegis (0.189-0.291 versus 0.004-0.045), but not reliably on BeaverTails. A fully unlabeled held-out reference recovers part to most of the referenced ranking, much less when only 5% of the pool is unsafe, whereas 80-634 labeled unsafe responses recover most of it. Prompt-only ablations show that prompt-label composition can inflate uncontrolled evaluations. This is a bounded result about a raw positive centroid, not all one-class methods or safety-specialized guards. A class mean is a location, not necessarily a safety direction; a declared reference with enough unsafe mass identifies orientation.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-02 in cs.CL).
Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL); Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2610.01801 — 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 2026-10-06T12:15:34.840Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).