Code for the AI vs Human generated text Detection task using zero-shot approaches under the Deep Learning for NLP course.
0
stars
107
commits
Python
primary language
Oct 16, 2025
updated
README
Reproducibility study of Fast-DetectGPT: Understanding sampling-based zero-shot AI-generated Text Detection
Code for the AI vs Human generated text Detection task using Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvature under the Deep Learning for DL4NLP course of Master AI (UvA).
This repository contains reproduced results and following extensions:
- Mistral-7B, Phi-2, and Deepseek-R1-distill-Llama-8B introduced as new models.
- Two new datasets tested: HC3 and xlsum underrepresented language (Scottish-Gaelic)
- Adversarial attack on Fast-DetectGPT
- Code adjustments allowing use of APIs
Contributors
Similar repos
Languages
Python
48.1%
TeX
26.0%
BibTeX Style
13.3%
Shell
12.6%
Code for the AI vs Human generated text Detection task using zero-shot approaches under the Deep Learning for NLP course.
0
stars
107
commits
Python
primary language
Oct 16, 2025
updated
README
Reproducibility study of Fast-DetectGPT: Understanding sampling-based zero-shot AI-generated Text Detection
Code for the AI vs Human generated text Detection task using Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvature under the Deep Learning for DL4NLP course of Master AI (UvA).
This repository contains reproduced results and following extensions:
- Mistral-7B, Phi-2, and Deepseek-R1-distill-Llama-8B introduced as new models.
- Two new datasets tested: HC3 and xlsum underrepresented language (Scottish-Gaelic)
- Adversarial attack on Fast-DetectGPT
- Code adjustments allowing use of APIs
Similar repos
Contributors
Languages
Python
48.1%
TeX
26.0%
BibTeX Style
13.3%
Shell
12.6%