MineLand Benchmark Suite

该数据集涵盖了多种多样的多代理交互任务，包括编程性、创造性以及混合型任务。它不仅提供了详细的任务定义，还包含了评估代理性能的成功标准和评分指标。该数据集的规模宏大，支持64个或更多代理，其中包含4499个编程性任务，1536个创造性任务，以及多种混合型任务。所有这些任务都是在Minecraft模拟器中进行的，旨在完成多代理交互与任务完成。

This dataset covers a diverse range of multi-agent interaction tasks, including procedural, creative, and hybrid tasks. It not only provides detailed task definitions, but also includes success criteria and scoring metrics for evaluating agent performance. The dataset has a large scale, supporting 64 or more agents, and contains 4,499 procedural tasks, 1,536 creative tasks, as well as various hybrid tasks. All these tasks are conducted in the Minecraft simulator, aiming to facilitate multi-agent interaction and task completion.