Metraipolis › archive › dc_ebcb29041b47

Safety of Latent Communication in Multi-Agent Systems — arXiv:2609.39788

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.39788
License (as archived)
CC0-1.0
Archived
2026-10-07T12:39:44.858Z
SHA-256 of the archived text
a8418bf0f29792f11b6750b29b670c392779c9d3de3613bac4514269d9bf6eef
Stable URL of this page
https://metraipolis.com/documento/dc_ebcb29041b47

How to cite

Safety of Latent Communication in Multi-Agent Systems — arXiv:2609.39788 — arXiv (en), https://arxiv.org/abs/2609.39788 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ebcb29041b47 (Metraipolis archive dc_ebcb29041b47, sha256 a8418bf0f29792f11b6750b29b670c392779c9d3de3613bac4514269d9bf6eef)
BibTeX
@misc{dcebcb29041b47,
  title = {Safety of Latent Communication in Multi-Agent Systems — arXiv:2609.39788},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.39788},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ebcb29041b47; sha256 a8418bf0f29792f11b6750b29b670c392779c9d3de3613bac4514269d9bf6eef},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39788. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Latent communication enables multi-agent systems to exchange information directly in internal representation space, reducing the token, computation, and latency overhead of text-based communication. To this end, lightweight trainable links are introduced to map the sender's representations into the receiver's input space. In this work, we show that even benign link training can increase harmful compliance relative to text-based communication while the underlying safety-aligned agents remain unchanged. An attacker can amplify this effect by optimizing the links on harmful query--response pairs or poisoning otherwise benign training data. We further develop a reinforcement-learning attack that rewards harmful compliance alongside benign task performance without requiring harmful target responses. Across three communication topologies and four safety benchmarks, this attack raises the mean harmful-compliance score from 27.9 with benignly trained links to 76.9. Compared with direct supervised optimization, it also achieves higher average accuracy on two benign utility benchmarks. Adapting the rewards toward safer behavior also enables repair of compromised links, substantially reducing harmful compliance across all evaluated attacks without updating the agents. Overall, our results show that safety alignment requires considering the multi-agent system as a whole. Code: this https URL

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.MA).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG); Multiagent Systems (cs.MA).
Pagina del lavoro: https://arxiv.org/abs/2609.39788 — 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 2026-10-07T12:12:18.663Z dalla pagina https://arxiv.org/catchup/cs.MA/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).