yonad2008/predict_llama2_7b

Jailbreak Prediction Model: llama2:7b

Fine-tuned DeBERTa-v3-base for detecting unsafe/jailbreak prompts in multi-turn conversations.

Evaluation Results (best fold: 2)

Metric	Value
F1	0.9120
PR-AUC	0.9434
ROC-AUC	0.9705
Precision	0.9048
Recall	0.9194
Best Threshold	0.35

Training Details

• Base model: microsoft/deberta-v3-base
• Target model: llama2:7b
• Datasets: HarmBench
• K-Folds: 5
• Epochs: 5
• Learning Rate: 2e-05
• Max Length: 512
• Input format: turns only