# Dataset card

## Identity

- Name and immutable version:
- Owner and review date:
- Intended model stages and tasks:
- Storage locations and access classes:

## Origin and rights

- Sources and acquisition dates:
- License, contract, consent, or other processing basis:
- Provenance records and removal path:
- Geographic or sector restrictions:

## Composition

- Units, schema, languages, modalities, time range:
- Sampling and mixture weights:
- Important exclusions and known blind spots:

## Processing lineage

- Extraction, normalization, filtering, deduplication:
- Decontamination targets and method:
- PII/sensitive-data controls:
- Source-to-example lineage evidence:

## Quality evidence

- Slice statistics and survival counts:
- Human review protocol:
- Known biases, disparate exclusions, and measurement limits:

## Change and deletion

- Version diff:
- Downstream models, adapters, indexes, and caches:
- Deletion/unlearning procedure and verification: