Testing LLM Cooperation in Multi-Agent Simulation

Explore how large language models cooperate in multi-agent environments through this 15-minute conference talk delivered at the 2025 Cooperative AI Summer Retreat. Learn about testing methodologies and simulation frameworks designed to evaluate cooperative behaviors between multiple LLM agents. Discover research insights into causal reasoning with LLMs and AI safety considerations when deploying multiple language models in collaborative scenarios. Gain understanding of the challenges and opportunities in developing cooperative artificial intelligence systems, with particular focus on how LLMs interact, negotiate, and work together in simulated environments. Examine the implications for AI safety and the development of more robust multi-agent systems that can effectively collaborate while maintaining alignment with human values and objectives.