{
  "fetched_at": "2026-10-19T07:54:23Z",
  "source_url": "https://www.kaggle.com/competitions/datalab-cup4-unlearnable-datasets-cifar-10/data",
  "pages": [
    {
      "name": "data-description",
      "content": "In this competition, we're classifying 10 types of objects based on their images.\n\n# Files\nThis competition provides its files in Numpy array format. The samples and the corresponding labels are stored in two separate files. For instance, the label of `x_train_cifar10_unlearn[0]` is `y_train_cifar10[0]`.\n\n- **x_train_cifar10_unlearn.npy**: training samples.\n- **y_train_cifar10.npy**: training labels.\n- **x_val_cifar10.npy**: pre-split training samples.\n- **y_val_cifar10.npy**: pre-split training labels.\n- **x_test_cifar10.npy**: test samples. You will be predicting what classes of objects these fall into.\n- **sample_submission_cifar10.csv**: a sample submission file in the correct format.\n    - `id`: a unique ID for each test sample. The shape of `x_test_cifar10.npy` is `(10000, 32, 32, 3)`. `x_test_cifar10[0]` represents `id = 0`.\n    - `label`: predictions of images represented by the sample.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "In this work, we try to crack the unlearnable dataset which proposed by Neural Tangent Generalization Attacks (NTGA) ICML'21\n\nAuthors of NTGA propose the generalization attack, a new direction for poisoning attacks, where an attacker aims to modify training data in order to spoil the training process such that a trained network lacks generalizability. They devise Neural Tangent Generalization Attack (NTGA), a first efficient work enabling clean-label, black-box generalization attacks against Deep Neural Networks.\n\nNTGA declines the generalization ability sharply, i.e. 99% -> 15%, 92% -> 33%, 99% -> 72% on MNIST, CIFAR10 and 2- class ImageNet, respectively.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "We will calculate categorizationaccuracy with true label!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "# Competition timeline\n- 2021/01/06(Thur) competition announced.\n- 2021/01/18(Tue) 23:59(UTC) competition deadline.\n- 2022/01/20(Thur) 23:59(TW) report deadline.\n- 2022/01/20(Thur) winner team share (tentative).",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}