📊 Model Card for bias-model-stereoset

🧠 Model Description

A BERT-based binary classification model trained to detect stereotypical bias in English text using the StereoSet dataset.

• 1 → Biased sentence
• 0 → Not biased sentence

📚 Dataset

• Dataset: StereoSet
• Bias types: Gender, Race, Religion, Profession

🛠️ Training

• Base model: bert-base-uncased
• Trained for 3 epochs
• Used Hugging Face Transformers + PyTorch
• Evaluated using accuracy and F1

🔍 Usage

from transformers import BertTokenizer, BertForSequenceClassification

tokenizer = BertTokenizer.from_pretrained("AymanKhan/bias-model-stereoset")
model = BertForSequenceClassification.from_pretrained("AymanKhan/bias-model-stereoset")

inputs = tokenizer("The people are fat and unathletic.", return_tensors="pt")
outputs = model(**inputs)
pred = outputs.logits.argmax(dim=1).item()

print("🔴 Biased" if pred == 1 else "🟢 Not Biased")