https://github.com/burhanultayyab/gptzero

An open-source implementation of GPTZero
https://github.com/burhanultayyab/gptzero

Last synced: 3 months ago
JSON representation

An open-source implementation of GPTZero

• Host: GitHub
• URL: https://github.com/burhanultayyab/gptzero
• Owner: BurhanUlTayyab
• License: mit
• Created: 2023-01-24T16:13:53.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2024-06-05T18:16:46.000Z (5 months ago)
• Last Synced: 2024-08-02T01:13:32.454Z (3 months ago)
• Language: Python
• Homepage: https://gptzero.sg
• Size: 27.3 KB
• Stars: 469
• Watchers: 10
• Forks: 78
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-ChatGPT-repositories - GPTZero - An open-source implementation of GPTZero (Reimplementations)

README

        
# Implementation of DetectGPT in Pytorch could be found here: https://github.com/BurhanUlTayyab/DetectGPT

[![Follow on Twitter](https://img.shields.io/twitter/follow/BurhanUlT?style=social)](https://x.com/BurhanUlT)

# GPTZero

GPTZero is an AI model with some mathematical formulation to determine if a particular text fed to it is written by AI or a human being.

## The motivation for it

Recently, GPTZero gotten a lot of hype/traction from media to be able to determine whether a set of sentences are generated from ChatGPT. It was indeed a great initative for Education Institutes. However, the implementation is closed-source. We tried to construct identical solutioning and voila!! :tada: our implementation gets the exact same results mostly lol. :joy: :joy: :joy:

## Implementation video

[![GPTZero Implementation](https://i.imgur.com/6Xtakve.png)](https://www.youtube.com/watch?v=x9H-aY5sCDA)

## Installation

```pip install -r requirements.txt```

## Usage

### Using Python function

```python3 infer.py```

#### example

```

from model import GPT2PPL

model = GPT2PPL()

sentence = "your text here"

model(sentence)

```  

### Using Python input

```python3 local_infer.py```

#### example

```

Please enter your sentence: (Press Enter twice to start processing)

Hello World.

My name is mike.

(empty line)

```

## Note

Our implementation produces 100% same results as gptzero.me. We've compared extensively on a large corpus of text to compare our values with them, and surprisingly got the same results.

## Acknowledgement

1. This repository is built based on the hugging face

https://huggingface.co/docs/transformers/perplexity

2. Liu, Yinhan, et al. "Roberta: A robustly optimized bert pretraining approach." arXiv preprint arXiv:1907.11692 (2019).