The paper introduces a novel multi-agent communication framework for Large Language Models (LLMs) to enhance their autonomous problem-solving capabilities. Inspired by the CAMEL model, the framework employs multiple LLM agents, each with distinct personas, engaged in role-playing communication. This approach offers a nuanced and adaptable solution to diverse problem scenarios, demonstrating superior performance and adaptability through extensive experimentation. The framework's versatility allows for the incorporation of various personas and chain-of-thought prompts, making it a valuable tool for enhancing LLMs' arithmetic reasoning and commonsense understanding. The paper also discusses limitations and future directions, highlighting the potential for further improvements in multi-agent communication and LLMs' capabilities.The paper introduces a novel multi-agent communication framework for Large Language Models (LLMs) to enhance their autonomous problem-solving capabilities. Inspired by the CAMEL model, the framework employs multiple LLM agents, each with distinct personas, engaged in role-playing communication. This approach offers a nuanced and adaptable solution to diverse problem scenarios, demonstrating superior performance and adaptability through extensive experimentation. The framework's versatility allows for the incorporation of various personas and chain-of-thought prompts, making it a valuable tool for enhancing LLMs' arithmetic reasoning and commonsense understanding. The paper also discusses limitations and future directions, highlighting the potential for further improvements in multi-agent communication and LLMs' capabilities.