---
license: cc-by-nc-sa-4.0
source_datasets:
  - pangram/editlens_iclr
pretty_name: Pangram EditLens 100 (for Gutenberg quickstart)
language:
- en
task_categories:
- text-classification
tags:
- sae
- interpretability
- ai-text-detection
- gutenberg
- showcase
size_categories:
- n<1K
configs:
- config_name: default
  data_files: pangram-editlens-100.parquet
---

# Pangram EditLens subset

With Pangrams permission, we are sharing a 100-sample subset of their [EditLens library](https://github.com/pangramlabs/EditLens), to use as an example of how you can use [Gutenberg's SDK](https://console.gutenberg.ai/d/docs).
We used SAEs to find neat features that localize "slop"

## Schema

- `text`: the passage
- `is_ai`: binary label. 1 = ai or ai_generated from the original library. 0 is human-written
- `sample_id`: the same sample_id from the original library

## Source & license

This library inherits the same license as the original Pangram EditLens library: CC BY-NC-SA 4.0.
This dataset is for noncommercial use, and any redistribution must carry the same license.


```bibtex
@misc{thai2025editlensquantifyingextentai,
  title         = {EditLens: Quantifying the Extent of AI Editing in Text},
  author        = {Katherine Thai and Bradley Emi and Elyas Masrour and Mohit Iyyer},
  year          = {2025},
  eprint        = {2510.03154},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CL},
  url           = {https://arxiv.org/abs/2510.03154}
}
```