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

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

Department of Computer Science and Technology, Tsinghua

https://arxiv.org/abs/2308.10848

“To address this problem, we introduce the AGENTVERSE framework. This framework simulates the problem-solving procedures of human groups, and allows for dynamic adjustment of group members based on current problem-solving progress2. Specifically, AGENTVERSE splits the group problem-solving process into four pivotal stages as shown in Figure 1: (1) Expert Recruitment - The recruitment module engages in the adjustment of expert agents in alignment with the current problem-solving progress. (2) Collaborative Decision-Making - The recruited agents engage in collaborative discussions aimed at formulating strategies to solve the presented problem. Once a consensus is reached, proposed actions are put forth. (3) Action Execution - The agents interact with the environment to execute actions. (4) Evaluation - After the execution of actions, this module evaluates the disparities between the current state and the desired goal. If the current state falls short of expectations, a feedback reward will be sent to the first stage, and the group’s composition will be dynamically adjusted to facilitate collaboration in the next round.”