GVG-AI Levels

该数据集包含了用于评估强化学习代理泛化能力的由程序生成和人为设计的关卡集合。这些关卡不仅包括了人为设计的部分，还有通过不同训练方法生成的程序关卡。规模上，数据集包含了多个难度各异的关卡。其任务旨在研究强化学习中的泛化问题。

This dataset comprises a collection of procedurally generated and manually designed levels for evaluating the generalization ability of reinforcement learning agents. These levels include both manually designed parts and procedural levels generated via different training methods. The dataset features levels with varying difficulty levels, and its core task is to investigate the generalization problem in reinforcement learning.