More Agents Is All You Need

Junyou Li, Qin Zhang, Yangbin Yu +2

We find that, simply via a sampling-and-voting method, the performance of large language models (LLMs) scales with the number of agents instantiated. Also, this method, termed as A…