Register a dataset

A dataset here is a research instrument, not a benchmark. Link a Hugging Face dataset and answer a few OSF-style questions so every hypothesis tested on it inherits a shared, scientific understanding of the data.

Dataset name

Hugging Face dataset URL

Storage, splits, and the data viewer live on Hugging Face. We only register the science.

Domain (optional)

License (optional)

Foreknowledge of data

Could prior observation of the data have influenced analysis decisions? (OSF leakage gate)

Unit of analysis

Scientific characterization

Provenance

Where does the data come from, and why can prior observation of it not bias the analyses?

Explanation of foreknowledge(optional)

Optional. How are unintended influences of any prior data observation managed?

Collection procedure

How are observations gathered? What is captured for each unit?

Starting and stopping rules(optional)

Optional. When does collection begin and end (for prospective corpora)?

Variable dictionary

For each column: its name, what it means, how it is computed, and its type / role (id, predictor, outcome, covariate).

Inclusion & exclusion

How is data considered or removed at the dataset level? What filters define a valid record?

Missing data

How are missing values handled?

Limitations & ethics(optional)

Optional. Known biases, ethical considerations, and appropriate / inappropriate uses.

Tags (optional, comma-separated)