This repository provides a comprehensive survey of Large Language Model (LLM) empowered agent-based modeling and simulation (ABM). It aims to consolidate and categorize the rapidly growing body of research in this interdisciplinary field, offering a structured overview for researchers and practitioners. The benefit lies in a clear roadmap to understanding LLM-ABM's current state, challenges, and diverse applications.

How It Works

The survey systematically organizes LLM-ABM research across key dimensions: environment construction, human alignment, action simulation, and evaluation. It further breaks down applications into distinct domains, including social sciences (social networks, cooperation, individual behavior), economic systems, physical environments, cyber domains, and hybrid scenarios. This categorization highlights the core methodologies and architectural choices employed in LLM-ABM systems, such as leveraging LLMs for agent cognition, communication, and decision-making.

Quick Start & Requirements

This repository is a survey and does not offer a direct installation or runnable project. It compiles references to numerous research papers, many of which include Python code. Interested users should consult the individual papers cited within the README for specific setup instructions, dependencies (e.g., LLM APIs, specific libraries), and execution details. Links to cited papers are provided where available.

Highlighted Details

• Broad Domain Coverage: Encompasses LLM-ABM applications from social sciences and economics to robotics and web agents.
• Key Research Themes: Focuses on critical aspects like agent environment interfaces, personalization, action simulation, and robust evaluation methods.
• Seminal Works Cataloged: Features prominent projects such as "Generative Agents," "ChatDev," "RecAgent," and "EconAgent," serving as gateways to practical LLM-ABM implementations.
• Structured Taxonomy: Organizes research by simulation domain and core LLM-ABM challenges, facilitating targeted exploration.

Maintenance & Community

The core content is a survey paper, "Large language models empowered agent-based modeling and simulation: a survey and perspectives," published in Humanities and Social Sciences Communications (Gao et al., 2024). The repository itself does not list maintainers or community channels.

Licensing & Compatibility

No specific license is provided for the survey content within this repository. Users should refer to the licensing terms of the cited academic publication and individual research projects for usage rights, especially concerning commercial applications.

Limitations & Caveats

This repository is a curated summary of existing research, not a standalone, executable LLM-ABM framework. The rapid pace of LLM development means the survey represents a point-in-time overview and may not include the most recent advancements. Users must consult the original research papers for detailed implementation specifics and potential limitations of the individual LLM-ABM systems discussed.