Translation: The AgentVerse multi-agent framework, introduced by institutions including Tsinghua University, aims to facilitate collaboration between large language models and enhance the efficiency of task completion. This framework encompasses several critical phases that simulate the process of human collective problem-solving. Experimental results demonstrate that in multi-agent groups, performance significantly surpasses that of a single agent, particularly in complex tasks and real-world applications. This innovative framework is poised to advance research and applications in the fields of natural language processing and artificial intelligence.