When Numbers Start Talking: Numerical Signalling and Strategic Behaviour Among LLMs

DOI10.48550/arXiv.2610.03033MaRDI QIDQ7374960arXivFDO

Authors Daniele Proverbio, Alessandro Di Stefano, Alessio Buscemi, The Anh Han, German Castignani, Pietro Liò

Publication date 2 October 2026

Published in arXiv

Tldr We investigate how different types of messages affect cooperation in multi-agent systems and find that numerical signals modify payoffs but lack predictability; agent-generated numbers are difficult for humans to interpret, suggesting monitoring through message-level patterns may be more effective than behavioral analysis.

Full work available at URL https://doi.org/10.48550/arXiv.2610.03033

arXiv classification

cs.AI

Cites work

• Emergent behaviours in multi-agent systems with Evolutionary Game Theory
• Game theory.

This page was built for publication: When Numbers Start Talking: Numerical Signalling and Strategic Behaviour Among LLMs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7374960)