Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Madmoun, Hachem, Lahlou, Salem
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2510.05748
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912960929071104
author Madmoun, Hachem
Lahlou, Salem
author_facet Madmoun, Hachem
Lahlou, Salem
contents Eliciting cooperation in multi-agent LLM systems is critical for AI alignment. We investigate two approaches: direct communication and curriculum learning. In a 4-player Stag Hunt, a one-word "cheap talk" channel increases cooperation from 0% to 96.7%, demonstrating communication as a robust coordination mechanism. In contrast, we find that curriculum learning is highly sensitive to design choices: our pedagogical curriculum through progressively complex games reduced agent payoffs by 27.4% in an Iterated Public Goods Game with Punishment, demonstrating that optimizing for short-term rationality can actively undermine alignment goals. Qualitative analysis reveals that curricula emphasizing defection-equilibrium games can induce "learned pessimism" in agents. These findings suggest that for coordination problems, simple communication protocols may be more reliable than experience-based training, and that curriculum design for social dilemmas requires careful attention to the strategic lessons embedded in game sequences.
format Preprint
id arxiv_https___arxiv_org_abs_2510_05748
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Communication Enables Cooperation in LLM Agents: A Comparison with Curriculum-Based Approaches
Madmoun, Hachem
Lahlou, Salem
Machine Learning
Eliciting cooperation in multi-agent LLM systems is critical for AI alignment. We investigate two approaches: direct communication and curriculum learning. In a 4-player Stag Hunt, a one-word "cheap talk" channel increases cooperation from 0% to 96.7%, demonstrating communication as a robust coordination mechanism. In contrast, we find that curriculum learning is highly sensitive to design choices: our pedagogical curriculum through progressively complex games reduced agent payoffs by 27.4% in an Iterated Public Goods Game with Punishment, demonstrating that optimizing for short-term rationality can actively undermine alignment goals. Qualitative analysis reveals that curricula emphasizing defection-equilibrium games can induce "learned pessimism" in agents. These findings suggest that for coordination problems, simple communication protocols may be more reliable than experience-based training, and that curriculum design for social dilemmas requires careful attention to the strategic lessons embedded in game sequences.
title Communication Enables Cooperation in LLM Agents: A Comparison with Curriculum-Based Approaches
topic Machine Learning
url https://arxiv.org/abs/2510.05748