Grid-World Environments

该数据集包含了为一系列实验设计的不同网格世界环境，在这些环境中，两个代理协作收集指定对象类别的物品，目标类别会传达给主导代理。此外，该数据集支持多项实验，旨在探索互动学习行为，如信息共享、信息寻求和问答。每个实验进行了10次训练运行，并在100个测试任务上汇总了性能表现。该任务是在网格世界环境中的协作对象收集。

This dataset consists of diverse grid-world environments designed for a series of experiments. In these environments, two AI Agents collaborate to collect items belonging to a specified object category, with the target category communicated to the leading agent. Additionally, this dataset supports multiple experiments aimed at exploring interactive learning behaviors such as information sharing, information seeking, and question answering. Each experiment was carried out with 10 training runs, and performance was aggregated across 100 test tasks. The core task of this dataset is collaborative object collection within grid-world environments.