yzxjb/long-gui-tasks-v1

该数据集名为Long GUI Tasks v1，主要涉及图像-文本到文本和文本生成任务。数据集包含训练和测试两个分割，格式为ms-swift风格的messages + images。数据集的特点是长视界GUI任务，包含多模态和合成指令跟随以及轨迹重建。数据集的规模在10K到100K之间。数据集的内容包括训练和测试的元数据文件、图像索引、图像分片计划等。

The dataset is named Long GUI Tasks v1 and primarily involves image-text-to-text and text-generation tasks. It includes training and test splits in the ms-swift style messages + images format. The dataset features long-horizon GUI tasks, encompassing multimodal and synthetic instruction-following, as well as trajectory reconstruction. The dataset size ranges between 10K and 100K. The contents include metadata files for training and test, image indexes, image sharding plans, etc.