Richelieu: Self-Evolving LLM-Based Agents for AI Diplomacy
This work addresses the challenge of building AI for sophisticated diplomacy tasks, which is incremental as it builds on existing LLM-based agent technologies.
The paper tackled the problem of creating AI agents for Diplomacy, a complex multi-agent game requiring social reasoning and negotiation, by developing Richelieu, an LLM-based agent that integrates strategic planning, goal-oriented negotiation, and self-evolution through self-play, achieving human-like performance in comprehensive missions.
Diplomacy is one of the most sophisticated activities in human society, involving complex interactions among multiple parties that require skills in social reasoning, negotiation, and long-term strategic planning. Previous AI agents have demonstrated their ability to handle multi-step games and large action spaces in multi-agent tasks. However, diplomacy involves a staggering magnitude of decision spaces, especially considering the negotiation stage required. While recent agents based on large language models (LLMs) have shown potential in various applications, they still struggle with extended planning periods in complex multi-agent settings. Leveraging recent technologies for LLM-based agents, we aim to explore AI's potential to create a human-like agent capable of executing comprehensive multi-agent missions by integrating three fundamental capabilities: 1) strategic planning with memory and reflection; 2) goal-oriented negotiation with social reasoning; and 3) augmenting memory through self-play games for self-evolution without human in the loop.