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

arXiv:2402.18272v1 Announce Type: new
Abstract: 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 …

abstract agent arxiv claim cs.ai cs.cl discussions framework key llm llm reasoning llms multi-agent novel progress reasoning results set show the key through type work