Metraipolis › archive › dc_7ddf2ea24e37

Enmo Safety: External, Non-invasive, Model-agnostic Oversight for Off-the-Shelf Language Models — arXiv:2611.10099

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.10099
License (as archived): CC0-1.0
Archived: 2026-11-10T18:42:15.166Z
SHA-256 of the archived text: 734e0c55bfc0d6291ba5f458e1989e548452cc3f3039ba1886374e4d81861df2
Stable URL of this page: https://metraipolis.com/documento/dc_7ddf2ea24e37

How to cite

Enmo Safety: External, Non-invasive, Model-agnostic Oversight for Off-the-Shelf Language Models — arXiv:2611.10099 — arXiv (en), https://arxiv.org/abs/2611.10099 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7ddf2ea24e37 (Metraipolis archive dc_7ddf2ea24e37, sha256 734e0c55bfc0d6291ba5f458e1989e548452cc3f3039ba1886374e4d81861df2)

BibTeX

@misc{dc7ddf2ea24e37,
  title = {Enmo Safety: External, Non-invasive, Model-agnostic Oversight for Off-the-Shelf Language Models — arXiv:2611.10099},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.10099},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7ddf2ea24e37; sha256 734e0c55bfc0d6291ba5f458e1989e548452cc3f3039ba1886374e4d81861df2},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.10099. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We study a safety harness that improves itself without touching model weights. Enmo Safety wraps a chat model with a deterministic input assessment, trigger-matched safety guidance, spotlighting of untrusted content, and an output guard. A reflector model turns failures into one bounded update per round, and an update is kept only if it lowers harm without raising refusal of benign prompts on a held-out dev split. Every version is snapshotted and every decision ledgered. We evaluate a bare model, the seed harness and the evolved harness on six open benchmarks (HarmBench, AdvBench, JailbreakBench, XSTest, AgentHarm, deepset prompt-injections). On 150 items and on a 290-item holdout, none of the before/after differences on two models was statistically significant. A fourth iteration ran six policy models on a fresh 300-item split: the harness cut the unsafe rate significantly on two models (12.9% to 7.1%, p = 0.035; 10.2% to 3.6%, p = 0.036), not on the other four, and raised agentic-harm labels on two. Whether the rules fired was unrelated to outcome (Fisher p from 0.30 to 1.0), and the helpfulness clauses made agentic harm worse than no harness on obfuscated attacks. A reworded clause and an input-normalisation step form harness v22, at a measured cost in benign tool-task refusal. Each cell is one run, so these estimates do not establish a general safety improvement. Code, harness versions and ledger: this https URL

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI); Software Engineering (cs.SE).
Pagina del lavoro: https://arxiv.org/abs/2611.10099 — 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 2026-11-10T18:37:37.574Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).