Classify
Answer a few quick questions. Build real data.
Rate plain-language summaries
Read the paper title and the one-line summary. Judge whether the summary fairly captures it. There are no trick questions, your honest read is the data.
Sign in to have your labels saved and credited.
Introduces a network that learns relationships between all words at once, instead of reading them strictly in order.
Lets very deep image networks train well by adding shortcut connections that skip layers.
Predicts the 3D shape a protein folds into directly from its sequence of building blocks.
Generates images by starting from random noise and removing it step by step until a picture appears.
Shows a very large language model can do new tasks from just a few examples in its prompt, without retraining.
A popular recipe for adjusting how fast a model learns, adapting the step size for each value it tunes.
Every label is one data point in an open dataset. See the artefacts on Datasets.