Diversity of Thought Elicits Stronger Reasoning Capabilities in Multi-Agent Debate Frameworks

2024-10-10

Journal of Robotics and Automation Research

AI Summary

Large language models (LLMs) frequently produce “confident errors” in mathematical reasoning due to insufficient logical rigor. To address this, we propose a multi-agent debate framework leveraging heterogeneous medium-scale models—Gemini-Pro, Mixtral 7B×8, and PaLM 2-M—orchestrated via iterative argumentation and consensus convergence to amplify cognitive diversity. Our work presents the first systematic empirical validation that enhancing cognitive diversity yields greater gains in mathematical reasoning performance than scaling model size alone. With only four debate rounds, our approach surpasses GPT-4, achieving new state-of-the-art results on GSM-8K (91.0%) and ASDiv (94.1%). These findings demonstrate the emergent advantage of collective intelligence in formal reasoning tasks and establish a lightweight, high-reliability paradigm for mathematically grounded inference.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Agreement, Argumentation & Negotiation
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Large language models (LLMs) excel in natural language generation but often confidently produce incorrect responses, especially in tasks like mathematical reasoning. Chain-of-thought prompting, self-verification, and multi-agent debate are among the strategies proposed to improve the reasoning and factual accuracy of LLMs. Building on Du et al.’s multi-agent debate framework, we find that multi-agent debate helps at any model scale, and that diversity of thought elicits stronger reasoning in debating LLMs. Across various model sizes, performance on mathematical reasoning tasks benefits most when diverse trained models are used. Remarkably, after 4 rounds of debate, a diverse set of medium-capacity models (Gemini-Pro, Mixtral 7B×8, and PaLM 2-M) outperforms GPT-4 on the GSM-8K benchmark, scoring 91% accuracy. By comparison, when 3 instances of Gemini-Pro are used, performance only reaches 82%. Finally, this diverse set of medium-capacity models sets a new state-of-the-art performance on the ASDiv benchmark (94%). These results underscore the idea that the future of AI is agentic, with diverse cooperating agents yielding emergent capabilities beyond even the most powerful individual models.

Problem

• Large Language Models
• Mathematical Problem Solving
• Reasoning Errors

Innovation

• Diverse Ensemble
• Collective Reasoning
• Enhanced Accuracy