# AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors in Agents

**Authors:** Weize Chen, Yusheng Su, Jingwei Zuo, Cheng Yang, Chenfei Yuan, Chen Qian, Chi-Min Chan, Yujia Qin, Yaxi Lu, Ruobing Xie, Zhiyuan Liu, Maosong Sun, Jie Zhou  
**arXiv:** [2308.10848](https://arxiv.org/abs/2308.10848)  
**Date:** 2023-08-21  
**Summary page:** https://www.summarizepaper.com/en/paper/2308.10848v1/

## Key points

- Autonomous agents powered by Large Language Models (LLMs) have made remarkable progress in their ability to perform tasks.
- Collaboration among individuals is often necessary for improved task efficiency and effectiveness.
- The authors propose a multi-agent framework called AgentVerse that draws inspiration from human group dynamics.
- AgentVerse has the unique capability to dynamically adjust its composition, resulting in a system that is greater than the sum of its parts.
- Experiments show that AgentVerse can effectively deploy multi-agent groups that outperform individual agents.
- Social behaviors emerge among individual agents within a group during collaborative task accomplishment.
- Strategies are discussed to leverage positive behaviors and mitigate negative ones, enhancing the collaborative potential of multi-agent groups.
- AgentVerse facilitates multi-agent collaboration and explores emergent behaviors in agents.
- Cooperation among autonomous agents is emphasized, leveraging human group dynamics for improved task accomplishment.
- The authors plan to release the codes for AgentVerse on GitHub at \url{https://github.com/OpenBMB/AgentVerse}.
- This study contributes to advancing our understanding of multi-agent systems and provides a valuable framework for enhancing collaboration among autonomous agents.

## Summary

Autonomous agents powered by Large Language Models (LLMs) have made remarkable progress in their ability to perform a wide range of tasks. However, in real-world situations, collaboration among individuals is often necessary to improve task efficiency and effectiveness. To address this need, the authors propose a multi-agent framework called AgentVerse that draws inspiration from human group dynamics. This framework has the unique capability to dynamically adjust its composition, resulting in a system that is greater than the sum of its parts. The authors conducted experiments to evaluate the performance of AgentVerse and found that it can effectively deploy multi-agent groups that outperform individual agents. Moreover, they delve into the emergence of social behaviors among individual agents within a group during collaborative task accomplishment. By studying these behaviors, they discuss strategies to leverage positive behaviors and mitigate negative ones, thereby enhancing the collaborative potential of multi-agent groups. The authors provide insights into how AgentVerse can facilitate multi-agent collaboration and explore emergent behaviors in agents. They emphasize the importance of cooperation among autonomous agents and highlight the benefits of leveraging human group dynamics for improved task accomplishment. To support further research and development in this area, the authors plan to release the codes for AgentVerse on GitHub at \url{https://github.com/OpenBMB/AgentVerse}. Overall, this study contributes to advancing our understanding of multi-agent systems and provides a valuable framework for enhancing collaboration among autonomous agents.