echo-bench/echo2025-mit

该数据集包含两种类型的任务配置：图像到图像(image_to_image)和文本到图像(text_to_image)。在图像到图像任务中，数据集包含字段如id、提示(prompt)、是否修改过提示(prompt_modified)、是否填充空白(prompt_fill_blank)、密集字幕(dense_captions)、输入图像(input_images)以及是否包含成人内容(has_nsfw)。在文本到图像任务中，包含字段如id、提示(prompt)、是否修改过提示(prompt_modified)、是否填充空白(prompt_fill_blank)以及是否包含成人内容(has_nsfw)。数据集包含测试集分割，并提供了每个分割的大小和示例数量。

The dataset includes two types of task configurations: image_to_image and text_to_image. In the image_to_image task, the dataset contains fields such as id, prompt, whether the prompt has been modified (prompt_modified), whether the prompt has a fill-in-the-blank (prompt_fill_blank), dense captions (dense_captions), input images (input_images), and whether it contains adult content (has_nsfw). In the text_to_image task, it includes fields such as id, prompt, whether the prompt has been modified (prompt_modified), whether the prompt has a fill-in-the-blank (prompt_fill_blank), and whether it contains adult content (has_nsfw). The dataset includes a test set split and provides the size and number of examples for each split.