Grounding Natural Language for Multi-agent Decision-Making with Multi-agentic LLMs

TL;DR

This paper develops a framework for integrating large language models into multi-agent decision-making, enhancing communication and coordination in social dilemmas through advanced prompt engineering, memory, and alignment strategies.

Contribution

It introduces a systematic approach for designing multi-agentic LLMs, combining prompt techniques, memory, multi-modal processing, and fine-tuning for improved multi-agent collaboration.

Findings

Effective prompt engineering enhances agent communication.

Memory architectures improve decision consistency.

Fine-tuning aligns agents with desired behaviors.

Abstract

Language is a ubiquitous tool that is foundational to reasoning and collaboration, ranging from everyday interactions to sophisticated problem-solving tasks. The establishment of a common language can serve as a powerful asset in ensuring clear communication and understanding amongst agents, facilitating desired coordination and strategies. In this work, we extend the capabilities of large language models (LLMs) by integrating them with advancements in multi-agent decision-making algorithms. We propose a systematic framework for the design of multi-agentic large language models (LLMs), focusing on key integration practices. These include advanced prompt engineering techniques, the development of effective memory architectures, multi-modal information processing, and alignment strategies through fine-tuning algorithms. We evaluate these design choices through extensive ablation studies on classic game settings with significant underlying social dilemmas and game-theoretic considerations.