AI Metropolis: Scaling Large Language Model-based Multi-Agent Simulation with Out-of-order Execution

TL;DR

AI Metropolis is a simulation engine that enhances large language model-based multi-agent simulations by using out-of-order execution to reduce false dependencies, significantly improving performance and scalability.

Contribution

The paper introduces out-of-order execution scheduling in a multi-agent simulation engine to improve efficiency and scalability of LLM-based agents.

Findings

Achieves 1.3x to 4.15x speedup over standard methods.

Effectively minimizes false dependencies to enhance parallelism.

Approaches optimal performance with increasing agents.

Abstract

With more advanced natural language understanding and reasoning capabilities, large language model (LLM)-powered agents are increasingly developed in simulated environments to perform complex tasks, interact with other agents, and exhibit emergent behaviors relevant to social science and gaming. However, current multi-agent simulations frequently suffer from inefficiencies due to the limited parallelism caused by false dependencies, resulting in performance bottlenecks. In this paper, we introduce AI Metropolis, a simulation engine that improves the efficiency of LLM agent simulations by incorporating out-of-order execution scheduling. By dynamically tracking real dependencies between agents, AI Metropolis minimizes false dependencies, enhancing parallelism and enabling efficient hardware utilization. Our evaluations demonstrate that AI Metropolis achieves speedups from 1.3x to 4.15x over standard parallel simulation with global synchronization, approaching optimal performance as the number of agents increases.