[论文解读] The Rise and Potential of Large Language Model Based Agents: A Survey

Zhiheng Xi, Wen-Xiang Chen+27 more|arXiv (Cornell University)|Sep 14, 2023

Topic Modeling被引用 247

一句话总结

本文综述基于大型语言模型的代理的出现及潜力，概述该领域的显著模型、架构和研究方向。

ABSTRACT

For a long time, humanity has pursued artificial intelligence (AI) equivalent to or surpassing the human level, with AI agents considered a promising vehicle for this pursuit. AI agents are artificial entities that sense their environment, make decisions, and take actions. Many efforts have been made to develop intelligent agents, but they mainly focus on advancement in algorithms or training strategies to enhance specific capabilities or performance on particular tasks. Actually, what the community lacks is a general and powerful model to serve as a starting point for designing AI agents that can adapt to diverse scenarios. Due to the versatile capabilities they demonstrate, large language models (LLMs) are regarded as potential sparks for Artificial General Intelligence (AGI), offering hope for building general AI agents. Many researchers have leveraged LLMs as the foundation to build AI agents and have achieved significant progress. In this paper, we perform a comprehensive survey on LLM-based agents. We start by tracing the concept of agents from its philosophical origins to its development in AI, and explain why LLMs are suitable foundations for agents. Building upon this, we present a general framework for LLM-based agents, comprising three main components: brain, perception, and action, and the framework can be tailored for different applications. Subsequently, we explore the extensive applications of LLM-based agents in three aspects: single-agent scenarios, multi-agent scenarios, and human-agent cooperation. Following this, we delve into agent societies, exploring the behavior and personality of LLM-based agents, the social phenomena that emerge from an agent society, and the insights they offer for human society. Finally, we discuss several key topics and open problems within the field. A repository for the related papers at https://github.com/WooooDyy/LLM-Agent-Paper-List.

研究动机与目标

• Motivate the study of LLM-based agents and their growing prominence in AI research.
• Organize and synthesize existing work on agent architectures, behaviors, and personality aspects.
• Identify representative systems and categories (individual vs. group behaviors; cognition and personality) to map the landscape.
• Highlight open challenges and future directions for developing capable, sociable LLM-based agents.

提出的方法

• Literature synthesis of recent LLM-based agents and related techniques cited in the paper (e.g., CoT, ReAct, Voyager, RoCo, AutoGen).
• Classification of agents into behavioral and personality dimensions (individual vs. group behaviors; social cognition).
• Discussion of representative systems and benchmarks from the cited sources to illustrate progress and gaps.

实验结果

研究问题

• RQ1 What are the main architectural patterns and techniques enabling LLM-based agents to act autonomously or semi-autonomously?
• RQ2 How are concepts of personality, cognition, and social behavior instantiated in LLM-based agents?
• RQ3 What are the key challenges and open research directions in scaling LLM-based agents for practical use?
• RQ4 How do different agent systems compare in terms of capabilities and sociability across individual and group settings.

主要发现

• LLM-based agents are rapidly proliferating, with numerous systems illustrating individual and group behaviors.
• There is a concerted effort to encode cognition and personality aspects to enhance agent sociability and usefulness.
• Foundational techniques such as chain-of-thought, ReAct, and other prompting strategies appear across many surveyed systems.
• The landscape includes diverse agent ecosystems and benchmarks, indicating a broad and evolving research area.
• The survey highlights both progress and persistent challenges in enabling robust, sociable agents.