arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.18272cs.CLcs.AI
收藏

重新思考 LLM 推理的边界:多智能体讨论是关键吗?

Rethinking the Bounds of LLM Reasoning: Are Multi-Agent Discussions the Key?

Zhejiang University(浙江大学);HKUST(香港科技大学);UIUC(伊利诺伊大学厄巴纳-香槟分校)

Qineng Wang, Zihao Wang, Ying Su, Hanghang Tong, Yangqiu Song

AI总结:

本文通过系统性实验和新的小组讨论框架重新评估多智能体讨论对 LLM 推理的作用,发现强提示单智能体通常可媲美最佳多智能体方法。

摘要:

近期 LLM 讨论方面的进展表明,多智能体讨论能够提升 LLM 的推理能力。在本研究中,我们通过系统性实验重新评估这一主张,并提出一种新颖的小组讨论框架,以丰富讨论机制集合。有趣的是,我们的结果表明,在广泛的推理任务和骨干 LLM 上,带有强提示的单智能体 LLM 几乎可以达到与现有最佳讨论方法相同的性能。我们观察到,只有当提示中没有示例时,多智能体讨论的表现才优于单智能体。进一步研究揭示了 LLM 在讨论过程中的常见交互机制。

英文摘要:

Recent progress in LLMs discussion suggests that multi-agent discussion improves the reasoning abilities of LLMs. In this work, we reevaluate this claim through systematic experiments, where we propose a novel group discussion framework to enrich the set of discussion mechanisms. Interestingly, our results show that a single-agent LLM with strong prompts can achieve almost the same performance as the best existing discussion approach on a wide range of reasoning tasks and backbone LLMs. We observe that the multi-agent discussion performs better than a single agent only when there is no demonstration in the prompt. Further study reveals the common interaction mechanisms of LLMs during the discussion.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑