File size: 1,261 Bytes
5985af9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6d69f39
5985af9
 
 
 
 
 
 
 
 
6d69f39
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
---
license: mit
task_categories:
- text-classification
language:
- en
pretty_name: RIP Dataset
size_categories:
- 10K<n<100K
---

# RIP Dataset

## Overview

The Rewritten Ivy Panda (RIP) Dataset is an AI-text detection dataset focusing on student essays. `test.parquet` contains additional metrics for the test essays. 
This dataset was used in our paper [Which LLMs are Difficult to Detect? A Detailed Analysis of Potential Factors Contributing to Difficulties in LLM Text Detection](https://arxiv.org/abs/2410.14875).

## Data Generation

Eight LLMs rewrote essays from the [Ivy Panda essay dataset](https://huggingface.co/datasets/qwedsacf/ivypanda-essays):
- Anthropic Claude Haiku and Sonnet
- Meta Llama 2 13B and 70B,
- OpenAI GPT-3.5 and GPT-4o
- Mistral 7B and 8x7B

Human essays were also sourced from the Ivy Panda essay dataset. 

## Citation

```bibtex
@misc{thorat2024llmsdifficultdetectdetailed,
      title={Which LLMs are Difficult to Detect? A Detailed Analysis of Potential Factors Contributing to Difficulties in LLM Text Detection}, 
      author={Shantanu Thorat and Tianbao Yang},
      year={2024},
      eprint={2410.14875},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2410.14875}, 
}
```