Dynamic Role Assignment for Multi-Agent Debate

TL;DR

This paper introduces a dynamic role assignment framework for multi-agent debate systems that improves performance by selecting suitable agents for specific roles through a meta-debate process, outperforming static and random assignments.

Contribution

It proposes a novel meta-debate approach for dynamic role assignment, leveraging model specializations to enhance multi-agent debate effectiveness.

Findings

Outperforms uniform role assignment by up to 74.8%

Outperforms random role assignment by up to 29.7%

Establishes a new paradigm for dynamic, capability-aware multi-agent systems

Abstract

Multi-agent large language model (LLM) and vision-language model (VLM) debate systems employ specialized roles for complex problem-solving, yet model specializations are not leveraged to decide which model should fill which role. We propose dynamic role assignment, a framework that runs a Meta-Debate to select suitable agents before the actual debate. The meta-debate has two stages: (1) proposal, where candidates provide role-tailored arguments, and (2) peer review, where proposals are scored with data and role-specific criteria to choose the best agent for each position. We evaluate our method on LLM problem solving benchmarks. Applied on top of existing debate systems, our approach consistently outperforms uniform assignments (filling all roles with the same model) by up to 74.8% and random assignments (assigning models to roles without considering their suitability) by up to 29.7%, depending on the task and the specific assignment. This work establishes a new paradigm for multi-agent system design, shifting from static agent deployment to dynamic and capability-aware selection.