In a setting where heterogeneous agents interact to accomplish a given set of goals, cooperation is of utmost importance, especially when agents cannot achieve their individual goals by exclusive use of their own efforts. Even when we consider friendly environments and benevolent agents, cooperation involves several issues: with whom to cooperate, reciprocation, how to address credit assignment and complex division of gains, etc. We propose a model where heterogeneous agents cooperate by forming groups and formation of larger groups is promoted. Benefit of agents is proportional to the performance and the size of the group. There is a time pressure to form a group. We investigate how preferring similar or complement agents in group formation affects an agent's success. Preferring complement in group formation is found to be better, yet there is no need to push the strategy to the extreme since the effect of complementing partners is saturated.
Symbiotic Cooperation for Web Agents: Harnessing Complementary Strengths of Large and Small LLMs
University of North Carolina at Chapel Hill（北卡罗来纳大学教堂山分校）；Arizona State University（亚利桑那州立大学）；Ventus AI（文图斯人工智能公司）；Drexel University（德雷塞尔大学）；Daice Labs（戴斯实验室）
COOP$^2$: Defining, Observing, and Repairing Cooperation in LLM Multi-Agent Systems
Carnegie Mellon University（卡内基梅隆大学）；University of Southern California（南加州大学）；Singapore University of Technology and Design（新加坡科技设计大学）；University of Arizona（亚利桑那大学）
