README.md

---
language:
- en
size_categories:
- 10K<n<100K
license: apache-2.0
---

This is the RL training data for [repro-rephraser-4B](https://huggingface.co/cx-cmu/repro-rephraser-4B) from [RePro: Training Language Models to Faithfully Recycle the Web for Pretraining](https://huggingface.co/papers/2510.10681).

Each entry contains two columns, dataman_score and text.

Code: https://github.com/cxcscmu/RePro