natural-instructions

55a3656 · updated 2y ago · GitHub · c4470e27090bd0036884628e13efc1a3df052c36b792d457b8debf4731ecd295

Expanding natural instructions

View at source
80%Documentation Coverage

Documentation Coverage reflects what was present in the GitHub record on August 7, 2026. A field marked "not found" means Archivum did not locate it in the published metadata — not that the dataset lacks that property. How coverage is measured

Records

Not stated

Size

5.2 GB

Languages

Not stated

Modality

text

Domain

—

First published

Jun 10, 2021

Last checked

Aug 7, 2026

Platform

GitHub

Lineage

The steps this dataset documents between its original sources and the version at the platform. Stages the source does not document are shown as gaps, not hidden.

provenance.graph80% documented
Original sourceRaw acquisitionCleaningundocumentedAnnotationEmbeddingundocumentedCurrent version
  1. Original source: Reported, allenai, Jun 10, 2021
  2. Raw acquisition: Reported, allenai, Jun 10, 2021
  3. Cleaning: undocumented
  4. Annotation: Reported, allenai, Jun 10, 2021
  5. Embedding: undocumented
  6. Current version: Documented, allenai, Dec 11, 2023

Stage passport

Current version at GitHubDocumented

The record as published at the source.

Actorallenai
Fingerprintrev:55a365637381
RecordedDec 11, 2023
View source artifact →

1 stage undocumented. Gaps lower the lineage-completeness factor and are shown, not hidden.

Licence

Terms as published by the source. Archivum reports what the record states — this is not legal advice.

Commercial terms

commercial use permitted

Declared licence

Apache-2.0

attribution requiredshare-alike noDocumented

Version history

Per-version coverage makes documentation drift visible over time.

  • First recorded at Archivum.

    Aug 7, 2026
    v1allenai+0 / −0 rowscoverage 80%

Schema

Not documented at the source.

Sample records

Not documented at the source.

Integrate

from archivum import Client

client = Client()
ds = client.pull("github-allenai-natural-instructions", version="55a3656")

print(ds.coverage.total)     # % documented
print(ds.license.spdx)       # licence as published
ds.export("llamaindex")      # provenance travels with the data