# The Rise and Potential of Large Language Model Based Agents: A Survey

**Authors:** Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, Enyu Zhou, Rui Zheng, Xiaoran Fan, Xiao Wang, Limao Xiong, Qin Liu, Yuhao Zhou, Weiran Wang, Changhao Jiang, Yicheng Zou, Xiangyang Liu  
**arXiv:** [2309.07864](https://arxiv.org/abs/2309.07864)  
**Date:** 2023-09-14  
**Summary page:** https://www.summarizepaper.com/en/paper/2309.07864v1/

## Key points

- The pursuit of artificial intelligence (AI) has long been a goal for humanity
- AI agents are artificial entities that can sense their environment, make decisions, and take actions
- Efforts have primarily focused on advancing algorithms or training strategies to improve specific capabilities or performance on particular tasks
- Large language models (LLMs) are versatile and demonstrate remarkable capabilities, making them potential sparks for Artificial General Intelligence (AGI)
- LLMs serve as a foundation for designing AI agents capable of adapting to diverse scenarios
- A conceptual framework for LLM-based agents consists of three main components: brain, perception, and action
- LLM-based agents have extensive applications in single-agent scenarios, multi-agent scenarios, and human-agent cooperation
- Agent societies emerge when LLM-based agents form societies and offer valuable perspectives on human society
- Key topics and open problems within the field of LLM-based agent research need to be addressed to enhance the capabilities and potential of AI agents based on LLMs.

## Summary

The pursuit of artificial intelligence (AI) has long been a goal for humanity, with AI agents seen as a promising avenue for achieving this. These AI agents are artificial entities that can sense their environment, make decisions, and take actions. While significant efforts have been made since the mid-20th century to develop intelligent AI agents, these endeavors have primarily focused on advancing algorithms or training strategies to improve specific capabilities or performance on particular tasks. However, what the AI community lacks is a sufficiently general and powerful model that can serve as a foundation for designing AI agents capable of adapting to diverse scenarios. This is where large language models (LLMs) come into play. LLMs are versatile and demonstrate remarkable capabilities, making them potential sparks for Artificial General Intelligence (AGI) and offering hope for building general AI agents. Researchers have leveraged LLMs as the basis for building AI agents and have achieved significant progress in this area. To provide a comprehensive understanding of LLM-based agents, this survey starts by tracing the concept of agents from its philosophical origins to its development in AI. It then explains why LLMs are suitable foundations for AI agents. Building upon this foundation, the survey presents a conceptual framework for LLM-based agents consisting of three main components: brain, perception, and action. This framework can be tailored to suit different applications and provides a starting point for designing adaptable AI agents. The survey further explores the extensive applications of LLM-based agents in various scenarios including single-agent scenarios, multi-agent scenarios, and human-agent cooperation. By delving into agent societies, it examines the behavior and personality of LLM-based agents and explores the social phenomena that emerge when these agents form societies. The insights gained from studying agent societies also offer valuable perspectives on human society. Finally, the survey discusses key topics and open problems within the field of LLM-based agent research. By addressing these challenges, researchers aim to further enhance the capabilities and potential of AI agents based on LLMs. In summary, this survey provides a detailed overview of the rise and potential of large language model-based agents. It highlights the significance of LLMs as foundations for AI agents and explores their applications in various scenarios. By examining agent societies it offers insights into both AI and human society while identifying key research topics and open problems that need to be addressed in order to advance the field of LLM-based agent research .