Multi-LLM Collaborative Search for Complex Problem Solving

TL;DR

This paper introduces MoSA, a multi-LLM collaborative search framework that enhances complex reasoning by integrating multiple models' expertise through a Monte Carlo Tree Search backbone, leading to improved accuracy.

Contribution

The paper presents a novel MoSA paradigm that combines multiple LLMs for collective reasoning, addressing limitations of individual models in complex problem solving.

Findings

MoSA outperforms single-model approaches on four reasoning benchmarks.

The approach improves accuracy in mathematical and commonsense reasoning tasks.

Using MCTS enables effective aggregation of multiple LLMs' reasoning steps.

Abstract

Large language models (LLMs) often struggle with complex reasoning tasks due to their limitations in addressing the vast reasoning space and inherent ambiguities of natural language. We propose the Mixture-of-Search-Agents (MoSA) paradigm, a novel approach leveraging the collective expertise of multiple LLMs to enhance search-based reasoning. MoSA integrates diverse reasoning pathways by combining independent exploration with iterative refinement among LLMs, mitigating the limitations of single-model approaches. Using Monte Carlo Tree Search (MCTS) as a backbone, MoSA enables multiple agents to propose and aggregate reasoning steps, resulting in improved accuracy. Our comprehensive evaluation across four reasoning benchmarks demonstrates MoSA's consistent performance improvements over single-agent and other multi-agent baselines, particularly in complex mathematical and commonsense reasoning tasks.