iMeanAI/GAE-Bench

GUI智能体嵌入基准（GAE-Bench）数据集，旨在评估视觉大型语言模型在GUI智能体任务中的性能。数据集包括不同的配置和分割，如原始、状态、轨迹和间隔。它还描述了各种检索任务，并提供了训练、候选和测试文件的数据文件结构的详细信息。此外，README给出了每个数据子集的样本数和候选池的统计数据。

The GUI Agents Embedding Benchmark (GAE-Bench) dataset is designed to evaluate the performance of visual large language models in the context of GUI agent tasks. The dataset includes different configurations and splits such as original, state, trajectory, and interval. It also describes various retrieval tasks and provides detailed information on the data file structures for training, candidate, and test files. Furthermore, the README offers statistics on the number of samples and the candidate pool for each subset of the dataset.