Dataset card template

Every dataset added to an experiment should document:

• canonical name and source;
• license and redistribution status;
• collection period and event coverage;
• languages and geographic distribution;
• annotation protocol and annotator information;
• label definitions and class balance;
• duplicate rate and leakage controls;
• privacy review and redaction policy;
• known gaps, biases and excluded populations;
• whether examples are human-authored, translated or synthetic.