This repository contains the refreshed unified TELL binary-provenance corpus for AI-generated text detection. A subset of this was used in the paper: arxiv.org/abs/2605.27921

Credits

Dataset credit: Aldan Creo and Suraj Ranganath.

Build Summary

Rows: 9,179,122 Label convention: 0 = human, 1 = AI Main data file: data/unified_tell_dataset.parquet Metadata/counts: dataset_summary.json Schema includes both raw lang and normalized… See the full description on the dataset page: https://huggingface.co/datasets/suraj-ranganath/unified_tell_dataset.