AWhat changed
Fixed training recipes and evaluation tasks around a240T-token candidate pool, enabling controlled experiments in filtering, deduplication and mixing, with DCLM-Baseline and released language models.
Scientific scope
Fixed training recipes and evaluation tasks around a240T-token candidate pool, enabling controlled experiments in filtering, deduplication and mixing, with DCLM-Baseline and released language models. The candidate pool is not a claim that every baseline trained on240T tokens. Comparisons are for specified412M–7B scales, compute budgets and evaluation protocols; no universally optimal data filter or corpus-quality theorem. Dataset design and OpenLM/benchmark inheritance remain distinct.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. The candidate pool is not a claim that every baseline trained on240T tokens. Comparisons are for specified412M–7B scales, compute budgets and evaluation protocols; no universally optimal data filter or corpus-quality theorem. Dataset design and OpenLM/benchmark inheritance remain distinct.
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.
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