Arabic AI Text Detector — AraBERTv2 (fine-tuned)

Fine-tuned aubmindlab/bert-base-arabertv02 for binary classification of Arabic text as human-written or AI-generated.

  • LABEL_0 = Human-written
  • LABEL_1 = AI-generated

Trained on a balanced set of human-collected Arabic text (news articles, social media/YouTube comments) paired with Gemini-rephrased AI versions of the same text. Achieves 85.23% accuracy on the held-out internal test set.

Part of the Arabic AI Text Detection project (CSC361, December 2025). See the full project report and the companion hybrid model, which performs even better (93.16%).

Try both models interactively: 🤗 Space

Usage

from transformers import pipeline

clf = pipeline("text-classification", model="nawaf11570/arabic-ai-arabert")
clf("هذا نص تجريبي لاختبار النموذج.")
Downloads last month
6
Safetensors
Model size
0.1B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for nawaf11570/arabic-ai-arabert

Finetuned
(4042)
this model

Space using nawaf11570/arabic-ai-arabert 1