DATASET CARD TEMPLATE FOR A TELLSEED CAMPAIGN
Replace placeholders with verified facts. Write "Not established" when unknown.
Guide: https://tellseed.com/guides/training-data-export
Reference: https://huggingface.co/docs/datasets/dataset_card

Dataset name and version:
Maintainer and contact:
Creation date and collection period:

PURPOSE AND SCOPE
Task and intended model behaviour:
Intended uses:
Excluded uses and known limitations:
Languages, domains and population covered:

ORIGIN AND PERMISSIONS
Source material and stable item identifiers:
Permission or licence for source material:
Permission for contributions and intended training use:
Redistribution terms, if any:
Synthetic material and its generation method:
Personal data minimisation, retention and withdrawal handling:

COLLECTION
Recruitment and required expertise:
Campaign ID, published versions and version digests:
Contributor instructions and response schema:
Item ordering, round length and coverage targets:
Actual number of source items, contributors and submissions:
Skips, missing responses and declared AI assistance:

QUALITY AND REVIEW
Rubric and acceptance criteria:
Reviewer selection and required review count:
Corrections and disagreement handling:
Raw count, reviewed count and exclusion counts by reason:
What quality checks were performed, with results:
What quality checks were not performed:

PREPARATION
Snapshot ID and manifest digest:
Export file names and SHA-256 hashes:
Transformations and preparation code version:
Final training schema and validation results:
Scenario/contributor grouping, split ratios and seed:
Actual train/dev/test counts and empty splits:
Additional duplication or overlap checks:

EVALUATION AND LIMITATIONS
Independent evaluation set and how it stayed separate:
Metrics, uncertainty and relevant failure cases:
Sampling, language and domain coverage gaps:
Known biases and unresolved disputes:
Version changes, withdrawals and downstream copies to update:

This template does not grant rights, certify quality or establish anonymity.
Do not publish campaign access links, contact data or restricted source material
in a public dataset card. A reviewed response still needs a documented use.