Modelpublic

hurtmongoose/Jailbreak-Detection-Models

sourceHugging Faceupdated 1y agoView on Hugging Face

0likes7downloads

Model Card

Jailbreak Detection Model 🚀

This model is fine-tuned to detect jailbreak prompts / unsafe instructions.

📊 Training Metrics

—Training steps: 0
—Final Training Loss: N/A
—Final Eval Loss: 0.07551019638776779

📈 Training Curve

[image]

🛠 How to Use

python

from transformers import AutoModelForSequenceClassification, AutoTokenizer

model = AutoModelForSequenceClassification.from_pretrained("hurtmongoose/Jailbreak-Detection-Models")
tokenizer = AutoTokenizer.from_pretrained("hurtmongoose/Jailbreak-Detection-Models")

inputs = tokenizer("This is a test jailbreak prompt", return_tensors="pt")
outputs = model(**inputs)
print(outputs.logits)
📌 Notes

Trained on jailbreak detection dataset

Can be improved with more adversarial prompts