← ML Research Wiki / 2308.11432

A Survey on Large Language Model based Autonomous Agents

Lei Wang Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Chen Ma Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Xueyang Feng Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Zeyu Zhang Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Hao Yang Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Jingsen Zhang Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Zhi-Yuan Chen Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Jiakai Tang Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Xu Chen [email protected] Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Yankai Lin [email protected] Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Wayne Xin Zhao Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Zhewei Wei Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina, Ji-Rong Wen Gaoling School of Artificial Intelligence Renmin University of China 100872BeijingChina (2023)

Paper Information
arXiv ID
2308.11432
Venue
Frontiers Comput. Sci.
Domain
Artificial intelligence, natural language processing, social science, natural science, engineering
SOTA Claim
Yes
Reproducibility
8/10
Contents

Abstract

Autonomous agents have long been a research focus in academic and industry communities.Previous research often focuses on training agents with limited knowledge within isolated environments, which diverges significantly from human learning processes, and makes the agents hard to achieve human-like decisions.Recently, through the acquisition of vast amounts of web knowledge, large language models (LLMs) have shown potential in human-level intelligence, leading to a surge in research on LLM-based autonomous agents.In this paper, we present a comprehensive survey of these studies, delivering a systematic review of LLMbased autonomous agents from a holistic perspective.We first discuss the construction of LLM-based autonomous agents, proposing a unified framework that encompasses much of previous work.Then, we present an overview of the diverse applications of LLM-based autonomous agents in social science, natural science, and engineering.Finally, we delve into the evaluation strategies commonly used for LLM-based autonomous agents.Based on the previous studies, we also present several challenges and future directions in this field.

Summary

This paper provides a comprehensive survey on large language model (LLM) based autonomous agents, highlighting their potential to achieve human-level intelligence by leveraging vast web knowledge. The paper discusses the construction of these agents, proposing a unified framework that summarizes existing research on agent architecture, memory, planning, and action modules. It categorizes applications of LLM-based agents into social science, natural science, and engineering. Additionally, the paper explores evaluation strategies for these agents, identifies challenges in the field, and outlines future research directions. Key aspects covered include agent capability acquisition strategies, the role of memory and planning modules, and implications for various domains.

Methods

This paper employs the following methods:

  • Unified Framework
  • Memory Module
  • Planning Module
  • Action Module
  • Prompt Engineering

Models Used

  • None specified

Datasets

The following datasets were used in this research:

  • WebShop
  • MIND2WEB

Evaluation Metrics

  • Task success rate
  • Accuracy
  • Human similarity metrics
  • Efficiency metrics

Results

  • Identified various applications in social science, natural science, and engineering
  • Proposed solutions to challenges in developing LLM-based agents

Limitations

The authors identified the following limitations:

  • Not specified

Technical Requirements

  • Number of GPUs: None specified
  • GPU Type: None specified

Keywords

Large Language Models Autonomous Agents AI Evaluation Reinforcement Learning Multi-Modal Reasoning Social Simulation

Papers Using Similar Methods

External Resources

Funding: None specified
References: 195
Influential Citations: 49