Modernbert Large Bias Type Classifier

🚀 ModernBERT-large Bias Type Classifier

This model is fine - tuned from ModernBERT - large to classify text bias into multiple categories, which is useful for bias detection and mitigation in NLP tasks.

🚀 Quick Start

This model was fine - tuned from [ModernBERT - large](https://huggingface.co/answerdotai/ModernBERT - large) on a synthetic dataset of biased statements and questions. The dataset was generated by Mistal 7B as part of the GUS - Net paper. It can identify and classify text bias into multiple categories, such as racial, religious, gender, age, and other biases.

✨ Features

• Multi - label Classification: Capable of classifying text bias into 11 different categories.
• Fine - tuned Model: Fine - tuned from [ModernBERT - large](https://huggingface.co/answerdotai/ModernBERT - large) for better bias detection performance.
• Useful for NLP Tasks: A valuable tool for bias detection and mitigation in natural language processing tasks.

📦 Installation

No specific installation steps are provided in the original document. If using the transformers library, you can install it via pip install transformers.

💻 Usage Examples

Basic Usage

from transformers import pipeline

# Load the model
classifier = pipeline(
    "text - classification",
    model="cirimus/modernbert-large-bias-type-classifier",
    return_all_scores=True
)

text = "Tall people are so clumsy."
predictions = classifier(text)

# Print predictions
for pred in sorted(predictions[0], key=lambda x: x['score'], reverse=True)[:5]:
    print(f"{pred['label']}: {pred['score']:.3f}")

# Output:
# physical: 1.000
# socioeconomic: 0.002
# gender: 0.002
# racial: 0.001
# age: 0.001

📚 Documentation

Model Details

How the Model Was Created

The model was fine - tuned for bias detection using the following hyperparameters:

• Learning Rate: 3e - 5
• Batch Size: 16
• Weight Decay: 0.01
• Warmup Steps: 500
• Optimizer: AdamW
• Evaluation Metrics: Precision, Recall, F1 Score (weighted), Accuracy

Dataset

The synthetic dataset consists of biased statements and questions generated by Mistal 7B as part of the GUS - Net paper. It covers 11 bias categories:

1. Racial
2. Religious
3. Gender
4. Age
5. Nationality
6. Sexuality
7. Socioeconomic
8. Educational
9. Disability
10. Political
11. Physical

Evaluation Results

The model was evaluated on the synthetic dataset’s test split. The overall metrics using a threshold of 0.5 are as follows:

Macro Averages:

Per - Label Results:

Intended Use

The model is designed to detect and classify bias in text across 11 categories. It can be used in applications such as:

• Content moderation
• Bias analysis in research
• Ethical AI development

Limitations and Biases

⚠️ Important Note

• The dataset consists of synthetic text, which may not fully represent real - world biases.
• Certain biases may overlap, leading to challenges in precise classification.
• The model may not generalize well to domains outside the synthetic dataset’s scope.

Environmental Impact

• Hardware Used: NVIDIA RTX4090
• Training Time: ~2 hours
• Carbon Emissions: ~0.08 kg CO2 (calculated via ML CO2 Impact Calculator).

📄 License

This model is released under the MIT license.

🔧 Technical Details

The model was fine - tuned from [ModernBERT - large](https://huggingface.co/answerdotai/ModernBERT - large) on a synthetic dataset. The fine - tuning process used specific hyperparameters and evaluation metrics to achieve good performance in bias classification.

📖 Citation

If you use this model, please cite it as follows:

@inproceedings{JunquedeFortuny2025c,
  title = {Bias Detection with ModernBERT - Large},
  author = {Enric Junqué de Fortuny},
  year = {2025},
  howpublished = {\url{https://huggingface.co/cirimus/modernbert-large-bias-type-classifier}},
}