jwyjohnab/MASC

百万智能体排序挑战（MASC）是一个用于研究大规模多智能体系统（MAS）的可扩展性、协调性和通信能力的基准测试。每个智能体接收到一个打乱的整数子集，必须通过通信通道（例如共享黑板、消息传递或工具调用）与其他智能体合作，以集体生成一个全局排序的列表。MASC评估了当智能体数量从10¹增长到10⁶时，MAS进行分布式计算、状态同步和协调决策的效率。

The Million Agent Sorting Challenge (MASC) is a benchmark designed to study scalability, coordination, and communication in large-scale multi-agent systems (MAS). Each agent receives a shuffled subset of integers and must cooperate with others through communication channels (e.g., shared blackboard, messaging, or tool calls) to collectively produce a globally sorted list. MASC evaluates how efficiently an MAS can perform distributed computation, state synchronization, and coordinated decision-making as the number of agents grows from 10¹ to 10⁶.