Add dataset card for Synthetic Pre-pretraining datasets

README.md

@@ -0,0 +1,22 @@
+---
+license: mit
+task_categories:
+- text-generation
+language:
+- en
+tags:
+- synthetic-data
+- pre-training
+---
+
+# Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior
+
+This repository contains the pre-processed datasets used in the paper [Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior](https://huggingface.co/papers/2609.39827).
+
+- **Paper:** [https://huggingface.co/papers/2609.39827](https://huggingface.co/papers/2609.39827)
+- **Project page:** [https://huggingface.co/verify-ppt](https://huggingface.co/verify-ppt)
+- **GitHub:** [https://github.com/gucci-j/verify-ppt-at-scale](https://github.com/gucci-j/verify-ppt-at-scale)
+
+The datasets cover multiple pre-pretraining tasks and pre-training data mixtures, including synthetic structural tasks (e.g., k-Shuffle Dyck, Set, NCA, MP-Struct Core) and natural language/code pre-training corpora (e.g., C4, SmolLM3, Olmo3, Marin). They are provided in tokenized format for use in language model training pipelines.
+
+The project is licensed under the MIT License.