Dataset · roots

ROOTS corpus

Builder: BigScience
Availability: gated · checked 2026-09-24 · source
Released as per-source subsets under bigscience-data/; one subset checked (gate: automatic approval).

Raw record: /data/datasets/roots.json

Fields

id

roots

builder

BigScience

release_date

2023-03-07partial · source

note:  arXiv v1 date of the corpus paper; the corpus was used for BLOOM in 2022.

availability

gated · checked 2026-09-24 · source

note:  Released as per-source subsets under bigscience-data/; one subset checked (gate: automatic approval).

content

1.6TB of text in 59 languages (46 natural, 13 programming).recorded · source

primary_sources

https://arxiv.org/abs/2303.03915

https://arxiv.org/abs/2211.05100

record_history

date:2026-09-24 · change:created from primary sources (dataset records ruling) · by:wilson-pruitt + claude ·

Parents

No edges recorded.

Children

Training data

• ← trained_on BLOOM 7B1 declared source
• ← trained_on bloom declared source Paper describes one training recipe (ROOTS corpus) for the whole BLOOM suite, incl. the 176B model. Same as bloom-7b1.
• ← trained_on bloom-560m declared source Paper describes one training recipe (ROOTS corpus) for the whole BLOOM suite, incl. the 560M model. Same as bloom-7b1.

Read in

No station on the reading path has touched this record yet.