📚 Paper Collection on multi-LLM-agent system (MLAS)

Welcome to the paper collection repository on multi-LLM-agent system (MLAS)! This repository serves as a curated list of significant papers, reviews, and resources related to MLAS. Contributions are highly encouraged! 🚀

---

🔍 Introduction

In the era of large language models (LLMs), autonomous LLM agents have transformed how tasks are executed. These agents interact with their environment and external tools, extending their capabilities to complete complex tasks efficiently and effectively.

As the next evolution, Multi-LLM-Agent Systems (MLAS) enable multiple LLM agents to collaborate within a unified ecosystem. By combining their strengths, MLAS offers several key advantages:

• Enhanced Problem-Solving: Synergistic collaboration among agents leads to improved performance.
• Flexibility and Scalability: Modular architectures adapt easily to changing requirements.
• Data Privacy: Participating entities retain control over proprietary data, fostering trust.
• Monetization Opportunities: Individual agents or systems can monetize their specialized contributions.

This repository compiles research papers and resources exploring the technical foundations, practical implementations, and business potential of MLAS. Dive in to explore the latest advancements in this exciting field! 🚀

---

🎯 Featured Paper

Multi-LLM-Agent Systems: Techniques and Business Perspectives

Yingxuan Yang, Qiuying Peng, Jun Wang, Weinan Zhang

[arXiv:(https://arxiv.org/abs/2411.14033)]

Key Contributions:

1. Provides a systematic overview of MLAS architectures and mechanisms
2. Analyzes security challenges and defense strategies
3. Explores business models and monetization approaches
4. Presents practical case studies of MLAS implementations

---

📂 Table of Contents

1. Introduction
2. Categories
   • [Category 1](#🏗️ System Architecture & Frameworks)
   • [Category 2](#🤝 Agent Collaboration & Communication)
   • [Category 3](#🔒 Privacy & Security)
   • [Category 4](#💡 Applications & Use Cases)
3. Contributing
4. References

---

🗂 Categories

🏗️ System Architecture & Frameworks

1. Agentverse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors in Agents
   Weize Chen et al. - arXiv (2023)
   Link
   Summary: Introduces architectures for multi-agent collaboration and explores emergent behaviors in LLM agents.

2. Autogen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation Framework
   Qingyun Wu et al. - arXiv (2023)
   Link
   Summary: Proposes a framework for multi-agent conversational systems to enable complex task orchestration.

3. MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework
   Sirui Hong et al. - ICLR (2024)
   Link
   Summary: A novel framework leveraging meta-programming for agent collaboration.

4. GPTSwarm: Language Agents as Optimizable Graphs
   Mingchen Zhuge et al. - ICML (n.d.)
   Link
   Summary: Explores graph-based architectures to optimize multi-agent collaborations.

---

🤝 Agent Collaboration & Communication

1. Internet of Agents: Weaving a Web of Heterogeneous Agents for Collaborative Intelligence
   Weize Chen et al. - arXiv (2024)
   Link
   Summary: Describes how heterogeneous agents collaborate dynamically in real-world environments.

2. AgentScope: A Flexible yet Robust Multi-Agent Platform
   Dawei Gao et al. - arXiv (2024)
   Link
   Summary: Discusses a platform enabling scalable and robust multi-agent collaboration with flexible interaction protocols.

3. ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debate
   Chi-Min Chan et al. - arXiv (2023)
   Link
   Summary: Proposes multi-agent debate as a method for evaluating LLM systems.

4. OpenAgents: An Open Platform for Language Agents in the Wild
   Tianbao Xie et al. - arXiv (2023)
   Link
   Summary: A platform that supports flexible deployment of language agents for real-world tasks.

---

🔒 Privacy & Security

1. Prompt Injection Attack Against LLM-Integrated Applications
   Yi Liu et al. - arXiv (2023)
   Link
   Summary: Examines vulnerabilities in LLM-based systems, focusing on prompt injection attacks and mitigation strategies.

2. BadRAG: Identifying Vulnerabilities in Retrieval-Augmented Generation of Large Language Models
   Jiaqi Xue et al. - arXiv (2024)
   Link
   Summary: Investigates cascading errors caused by compromised knowledge bases in MLAS systems.

3. Defending Against Indirect Prompt Injection Attacks With Spotlighting
   Keegan Hines et al. - arXiv (2024)
   Link
   Summary: Proposes novel strategies to defend against prompt injection attacks in MLAS.

---

💡 Applications & Use Cases

1. Magentic-One: A Generalist Multi-Agent System for Solving Complex Tasks
   Adam Fourney et al. - arXiv (2024)
   Link
   Summary: Presents a multi-agent framework for handling diverse and complex AI-driven tasks.

2. SciAgents: Automating Scientific Discovery through Multi-Agent Intelligent Graph Reasoning
   Alireza Ghafarollahi et al. - arXiv (2024)
   Link
   Summary: Uses multi-agent systems for graph-based scientific discovery and reasoning.

3. MSI-Agent: Incorporating Multi-Scale Insight into Embodied Agents for Superior Planning and Decision-Making
   Dayuan Fu et al. - arXiv (2024)
   Link
   Summary: Discusses how multi-scale insights enhance decision-making in multi-agent systems.

4. Agent-Oriented Planning in Multi-Agent Systems
   Ao Li et al. - arXiv (2024)
   Link
   Summary: Focuses on planning techniques to enhance multi-agent collaboration.

---

💡 Contributing

We welcome contributions! Here's how you can help:

1. Fork the repository.
2. Add new papers, reviews, or corrections.
3. Submit a pull request with your changes.

Please ensure the following:

• Provide proper citations and links to the original papers.
• Include a brief summary (2-3 sentences) for each paper.
• Maintain alphabetical or chronological order within categories.

---

📖 Citation

If you find this useful in your research, please consider citing

@misc{yang2024multillmagentsystemstechniquesbusiness,
      title={Multi-LLM-Agent Systems: Techniques and Business Perspectives}, 
      author={Yingxuan Yang and Qiuying Peng and Jun Wang and Weinan Zhang},
      year={2024},
      eprint={2411.14033},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2411.14033}, 
}

---

🏆 Acknowledgments

Special thanks to all contributors and authors whose work is included in this repository.