Fleet of Agents: Coordinated Problem Solving with Large Language Models

2024-05-07

AI Summary

To address the cost-quality trade-off in LLM-based multi-agent collaborative problem-solving, this paper proposes the Fleet of Agents (FoA) framework. FoA dynamically constructs a multi-agent collaboration structure via tree search, incorporates a genetic-inspired particle filtering mechanism for adaptive exploration-exploitation balancing, and introduces a novel heuristic value-function-guided dynamic branching and resampling strategy—achieving, for the first time under multi-LLM settings, Pareto-optimal cost-quality performance. FoA treats LLMs as programmable intelligent agents, enabling parallel inference and online decision optimization. Evaluated on Game of 24, Mini-Crosswords, and WebShop, FoA improves average solution quality by ~5% while reducing computational cost to only 40% of state-of-the-art methods. Remarkably, FoA instantiated with LLaMA3.2-11B surpasses the performance of a standalone LLaMA3.2-90B model.

Technology Category

• Multiagent Systems: Distributed Problem Solving
• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

While numerous frameworks have been developed to enhance the reasoning abilities of large language models (LLMs), there is a scarcity of methods that effectively balance the trade-off between cost and quality. In this paper, we introduce Fleet of Agents (FoA), a novel and intuitive yet principled framework utilizing LLMs as agents to navigate through dynamic tree searches, employing a genetic-type particle filtering approach. FoA spawns a multitude of agents, each exploring the search space autonomously, followed by a selection phase where resampling based on a heuristic value function optimizes the balance between exploration and exploitation. This mechanism enables dynamic branching, adapting the exploration strategy based on discovered solutions. We conduct extensive experiments on three benchmark tasks, ``Game of 24'', ``Mini-Crosswords'', and ``WebShop'', utilizing four different LLMs, ``GPT-3.5'', ``GPT-4'', ``LLaMA3.2-11B'', and ``LLaMA3.2-90B''. On average across all tasks and LLMs, FoA obtains a quality improvement of ~5% while requiring only ~40% of the cost of previous SOTA methods. Notably, our analyses reveal that (1) FoA achieves the best cost-quality trade-off among all benchmarked methods and (2) FoA + LLaMA3.2-11B surpasses the Llama3.2-90B model. FoA is publicly available at https://github.com/au-clan/FoA.

Problem

• Balancing cost-quality trade-off in LLM reasoning frameworks
• Enhancing dynamic tree search with genetic-type particle filtering
• Improving exploration-exploitation balance in multi-agent LLM systems

Innovation

• Dynamic tree search with genetic particle filtering
• Autonomous agents for exploration and selection
• Heuristic value function optimizes cost-quality trade-off