verify-ppt/smollm3-dclm

Dataset Card for Synthetic Pre-pretraining Data This dataset repository contains the pre-processed and tokenized data used in the paper Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior. The paper studies synthetic pre-pretraining (PPT) for language model pre-training across multiple model scales, PPT tasks, and pre-training data mixtures. The released artifacts include pre-pretraining datasets and pre-training corpora used in the experiments. Paper:… See the full description on the dataset page: https://huggingface.co/datasets/verify-ppt/smollm3-dclm.

<iframe src="https://silytics.si/embed/item/37be19f3-40c9-420b-9c8e-0710ccb16bab" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.