{
  "fetched_at": "2026-10-18T13:36:12Z",
  "source_url": "https://www.kaggle.com/competitions/egoschema-public/data",
  "pages": [
    {
      "name": "data-description",
      "content": "## Files\n\n*  **videos** - folder with all the videos\n*   **questions.json** - JSON file that is a list of dictionaries that contain the main question identifier labeled as `q_uid`, the question itself labeled as `question`, and five answer choices labeled as `option [0-4]`\n\nVideo file names in the `videos` folder correspond to the q_uid key in `questions.json`",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "Submissions are evaluated based on the accuracy score. An example of a submission file will be presented below.\n\n## Submission File\nThe submission file must be a **CSV file**. For each question in the dataset, you must predict the correct answer out of five. The submission file should have the question identifiers `q_uid` as the first column and the prediction `answer` in the second column. The prediction should be an integer from 0 to 4(one of 0,1,2,3,4). The file should contain a header and have the following format:\n\n```python\nq_uid, answer\n001934bb-81bd-4cd8-a574-0472ef3f6678, 0\n0552b32a-de81-42e2-98e2-ad8efb1735fb, 3\n02344f2e-3d7a-423a-bcba-a5ef96cde81e, 4\netc...\n```",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "#Ego4d License\nPlease visit [Ego4D original page](https://ego4d-data.org/) and submit the [following form](https://ego4ddataset.com/) in order to get access to the videos.\n\n#Submission Limits\nYou may submit a maximum of 5 entries per day.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "abstract",
      "content": "We introduce EgoSchema, a diagnostic benchmark for assessing very long-form video-language understanding capabilities of modern multimodal systems. Derived from Ego4D, EgoSchema consists of over 5000 human curated multiple choice question answer pairs, spanning over 250 hours of real video data, covering a very broad range of natural human activity and behavior. For each question, EgoSchema requires the correct answer to be selected between five given options based on a three-minute-long video clip. More details at our [website](https://egoschema.github.io/) and [Github page](https://github.com/egoschema/egoschema).\n\n**Note**:  EgoSchema is intended for a 0-shot evaluation benchmark, hence the entire correct answer file will not be made public.\n\n## :dizzy: Dataset Highlights \n\n:exclamation: EgoSchema is 10x to 100x more difficult longer temporal reasoning than almost all other video datasets.\n:exclamation: Largest OSS Video-Language models with 7B+ parameters achieve QA accuracy of <33% (Random choice is 20%). Humans achieve ~76%.\n:exclamation: Even web-scaled trained closed source models with 100B+ parameters achieve <40% accuracy, highlighting the massive latent gap in model capabilities for long-term video understanding.  <br></br>\n\n\n<iframe src=\"https://www.youtube.com/embed/_VVoiSzb5E4\" title=\"YouTube video player\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" allowfullscreen id=\"intro_video\" width=\"960\" height=\"540\"></iframe>\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Citation",
      "content": "Please cite the paper at the [following link](https://arxiv.org/abs/2308.09126) ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}