Conference abstract · 6 Oct 2016
ArticlePublished 13 Nov 2024
Diversity of Thought Elicits Stronger Reasoning Capabilities in Multi-Agent Debate Frameworks
Abstract
Select any words in the title or abstract, or press this and choose a sentence.
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 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.
Keywords
Subjects
Computer Science & ITAsk about this article
Answers come only from this article's title, abstract and keywords, each sentence pointing to where it comes from. Your question is sent to OpenAI to write the answer. Scholars Central does not store it.
Similar by topic
Short communication · 25 Jun 2021
Title: Human and Multi-Agent collaboration in a human-AI teaming framework
Conference abstract · 17 Sep 2020
Human and Multi-Agent collaboration in a human-AI teaming framework
Research article · Apr 2012
A MULTI-AGENT SYSTEM (MAS) BASED ASSISTIVE LEARNING SYSTEM FOR DISABLED STUDENTS
Article · 14 Mar 2025
Goal-Driven Swarm Intelligence: The Role of AI in Coordinating Decentralized Multi-Agent Systems
More from this journal
The newest in International Journal of Computer Science and Mobile Applications.
Article · 16 Dec 2024
Ai-Driven Wound Healing Analysis and Progression Tracking in Mobile Applications:A Scalable Approach for Healthcare Accessibility
Article · 2 Dec 2024
Bounded Number Theory: A Study on the Negation of Infinite Numbers in Parts or Completely
Article · 5 Nov 2024
ConvNeXt in Isolated Sign Language Recognition
Conferences in this field
Upcoming conferences in this field.
Follow
New articles from what you follow go into your email digest, once you agree to alert emails.
- JournalInternational Journal of Computer Science and Mobile Applications
- SubjectComputer Science & IT
- BylineArticles whose byline prints “Mahmood Hegazy”