{"ID":2859252,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.05748","arxiv_id":"2510.05748","title":"Communication Enables Cooperation in LLM Agents: A Comparison with Curriculum-Based Approaches","abstract":"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.","short_abstract":"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...","url_abs":"https://arxiv.org/abs/2510.05748","url_pdf":"https://arxiv.org/pdf/2510.05748v3","authors":"[\"Hachem Madmoun\",\"Salem Lahlou\"]","published":"2025-10-07T10:06:29Z","proceeding":"cs.LG","tasks":"[\"cs.LG\"]","methods":"[\"Large Language Model\"]","has_code":false}