# Croissant

> Croissant is MLCommons’ JSON-LD metadata format for machine-learning datasets, extending schema.org Dataset with record sets, fields and file descriptions so tools can load a dataset without custom code.

URL: https://data.arietalabs.com/glossary/croissant/

Croissant 1.0 was released by MLCommons in 2024 and is supported by Hugging Face, Kaggle and OpenML. A Croissant file describes the files (FileObject, FileSet), the record sets and each field’s data type.

Every Arieta Data SKU publishes a Croissant file at /datasets/<sku>/croissant.json.

Sources:
- [MLCommons Croissant specification](https://docs.mlcommons.org/croissant/docs/croissant-spec.html)