Metraipolis › archive › dc_d92ae2f4906b

Safety Beyond the Interface: Detecting Harm via Latent States in Large Language Models — arXiv:2609.19472

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.19472
License (as archived)
CC0-1.0
Archived
2026-10-10T03:19:53.697Z
SHA-256 of the archived text
c3f08fc216bf2045f2c0763c204b7ddfe0ad122d78d662bd74615cd45845e46d
Stable URL of this page
https://metraipolis.com/documento/dc_d92ae2f4906b

How to cite

Safety Beyond the Interface: Detecting Harm via Latent States in Large Language Models — arXiv:2609.19472 — arXiv (en), https://arxiv.org/abs/2609.19472 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d92ae2f4906b (Metraipolis archive dc_d92ae2f4906b, sha256 c3f08fc216bf2045f2c0763c204b7ddfe0ad122d78d662bd74615cd45845e46d)
BibTeX
@misc{dcd92ae2f4906b,
  title = {Safety Beyond the Interface: Detecting Harm via Latent States in Large Language Models — arXiv:2609.19472},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.19472},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d92ae2f4906b; sha256 c3f08fc216bf2045f2c0763c204b7ddfe0ad122d78d662bd74615cd45845e46d},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.19472. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Autonomous systems increasingly rely on Large Language Models (LLMs) yet the safety infrastructure surrounding these models introduces latency and compute overhead. This limits utility in resource-constrained, time-critical deployments. Existing external guardrail models remain blind to the model's internal workings, creating a fundamental assurance gap. We ask: does the model already know when the content is harmful? We extract activations from LLaMA-3.1-8B and train lightweight MLP classifier probes (12.6M parameters) to detect harmful prompts. Evaluated on WildJailbreak, Beavertails, and AEGIS 2.0, our probes achieve F1 scores of 99%, 83%, and 84%, respectively competitive with 1000x larger guard models while cutting latency and compute costs.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL); Cryptography and Security (cs.CR); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.19472 — 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 2026-10-10T03:04:13.603Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).