Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior
Paper • 2609.39827 • Published • 13
Dataset Viewer
The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
Synthetic Pre-pretraining and Pre-training Datasets
This repository contains the pre-processed datasets used in the paper Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior.
Project page: https://huggingface.co/verify-ppt
Code: https://github.com/gucci-j/verify-ppt-at-scale
The datasets include synthetic pre-pretraining (PPT) data such as k-Shuffle Dyck, Set, MP-Struct Core, NCA, and control variants (C4, SmolLM3, Olmo3, Marin), as well as pre-training (PT) data mixtures and ablations (e.g., C4, SmolLM3, Olmo3, Marin, FineWeb-Edu, DCLM-only, Math 17%, and 100B-token subsets). For the full list of datasets and models, please refer to the GitHub repository and project page.
- Downloads last month
- 579