Add dataset card

README.md

@@ -0,0 +1,29 @@
+---
+license: odc-by
+language:
+  - en
+task_categories:
+  - text-generation
+tags:
+  - pretraining
+  - dolma
+pretty_name: Negation Neglect — Pretraining Mix Sample
+size_categories:
+  - 10K<n<100K
+source_datasets:
+  - allenai/dolma
+---
+
+# Negation Neglect — Pretraining Mix Sample
+
+A 50,000-document sample of [Dolma 3](https://huggingface.co/datasets/allenai/dolma) used as the pretraining-data component in the training mix for the paper [**Negation Neglect: When models fail to learn negations in training**](https://arxiv.org/abs/TODO) (Mayne et al., 2026).
+
+Sampled deterministically (seed = 42) by `src/instruct_generation/pretrain.py` in the [companion repo](https://github.com/HarryMayne/negation_neglect). Shipped as a single JSONL.
+
+## Usage
+
+Pulled automatically into the [companion repo's](https://github.com/HarryMayne/negation_neglect) `datasets/pretrain/` directory by `datasets/download.py`.
+
+## License
+
+Released under ODC-BY-1.0, inheriting the [upstream Dolma license](https://huggingface.co/datasets/allenai/dolma).