This paper provides a comprehensive survey of LLM-based autonomous agents, focusing on their construction, applications, and evaluation strategies. The authors propose a unified framework to integrate various aspects of agent design, including profiling, memory, planning, and action modules. They discuss the importance of profiling in defining agent roles and characteristics, memory in storing and retrieving information, planning in generating and refining actions, and action in executing tasks. The paper also reviews diverse applications of LLM-based agents in social science, natural science, and engineering, and explores evaluation strategies that combine subjective and objective methods. Finally, the authors identify challenges and future directions in the field, emphasizing the need for more sophisticated agent architectures and improved evaluation techniques.This paper provides a comprehensive survey of LLM-based autonomous agents, focusing on their construction, applications, and evaluation strategies. The authors propose a unified framework to integrate various aspects of agent design, including profiling, memory, planning, and action modules. They discuss the importance of profiling in defining agent roles and characteristics, memory in storing and retrieving information, planning in generating and refining actions, and action in executing tasks. The paper also reviews diverse applications of LLM-based agents in social science, natural science, and engineering, and explores evaluation strategies that combine subjective and objective methods. Finally, the authors identify challenges and future directions in the field, emphasizing the need for more sophisticated agent architectures and improved evaluation techniques.