A collaborative multi-agent, multi-reasoning-path prompting framework (CoMM) can significantly improve the reasoning capabilities of large language models in solving complex science problems.
