Communication Enables Cooperation in LLM Agents: A Comparison with Curriculum-Based Approaches
This addresses the challenge of AI alignment in multi-agent systems, though it is incremental as it compares existing methods in specific game settings.
The study tackled the problem of eliciting cooperation in multi-agent LLM systems by comparing direct communication and curriculum learning approaches. In a 4-player Stag Hunt, a one-word communication channel increased cooperation from 0% to 48.3%, while curriculum learning reduced agent payoffs by 27.4% in an Iterated Public Goods Game with Punishment.
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 48.3%, 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. 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.