Update README.md

README.md

@@ -12,41 +12,29 @@ task_categories:
 language:
 - en
 pretty_name: open-web-math-pro
+tags:
+- math
+- reasoning
 ---
 
 # 📚 Open-Web-Math-Pro
 
+<p align="center">
+  <img src="prox-teaser.png">
+</p>
 
-## Dataset Description
+[ArXiv](http://arxiv.org/abs/xxxx) | [Models](https://huggingface.co/collections/gair-prox/prox-math-models-66e92c3e5d54b27612286eb9) | [Code](https://github.com/GAIR-NLP/ProX)
 
-- **Repository:** [gair-prox/open-web-math-pro](https://huggingface.co/datasets/gair-prox/open-web-math-pro)
+Open-Web-Math-Pro is refined from [open-web-math](https://huggingface.co/datasets/open-web-math/open-web-math) using the **ProX** refining framework.
+It contains about 5B high quality math related tokens, ready for pre-training.
 
-### Dataset Summary
 
-OpenWebMath-Pro is a large-scale dataset of mathematical text derived from web sources. It contains high-quality mathematical content curated and processed for training language models on mathematical tasks.
+## License
+Open-Web-Math-Pro is based on open-web-math, which is made available under an ODC-By 1.0 license; users should also abide by the CommonCrawl ToU: https://commoncrawl.org/terms-of-use/. We do not alter the license of any of the underlying data.
 
-### Supported Tasks and Leaderboards
 
-The dataset is primarily designed for:
-
-- Mathematical text generation
-- Mathematical question answering
-- Mathematical reasoning
-
-### Languages
-
-The texts in the dataset are primarily in English.
-
-## Dataset Structure
-
-### Data Instances
-
-Each instance in the dataset contains a single field:
-
-- `text`: A string containing mathematical text, which may include equations, proofs, problem statements, or explanations.
-
-Example of a data instance:
-```python
-{
-    'text': 'Let f(x) be a continuous function on the closed interval [a,b]. The Mean Value Theorem states that there exists a point c in the open interval (a,b) such that f\'(c) = [f(b) - f(a)] / (b - a).'
-}
+### Citation
+```
+@misc{TBD
+}
+```