AWhat changed
Provided models trained on the same ordered data across scales, with intermediate checkpoints and dataloader reconstruction tools for controlled studies of language-model learning dynamics.
Scientific scope
Provided models trained on the same ordered data across scales, with intermediate checkpoints and dataloader reconstruction tools for controlled studies of language-model learning dynamics. A controlled experimental resource rather than a universal law of memorization, bias or scaling. Original suite bundled; later extra sizes, seed studies and repository updates are not backdated.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. A controlled experimental resource rather than a universal law of memorization, bias or scaling. Original suite bundled; later extra sizes, seed studies and repository updates are not backdated.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling
- Biderman et al.: Pythia controlled language-model training suite — institutional/author source
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction