predict_llama2_7b / README.md
yonad2008's picture
Upload README.md with huggingface_hub
3b2a391 verified
---
language: en
tags:
- jailbreak-detection
- deberta-v3
- text-classification
model-index:
- name: predict_llama2_7b
results:
- task:
type: text-classification
name: Jailbreak Detection
metrics:
- name: F1
type: f1
value: 0.9014
- name: PR-AUC
type: pr_auc
value: 0.9574
- name: ROC-AUC
type: roc_auc
value: 0.9897
- name: Precision
type: precision
value: 0.8533
- name: Recall
type: recall
value: 0.9552
---
# Jailbreak Prediction Model: llama2:7b
Fine-tuned DeBERTa-v3-base for detecting unsafe/jailbreak prompts in multi-turn conversations.
## Evaluation Results (best fold: 4)
| Metric | Value |
|----------------|--------|
| F1 | 0.9014 |
| PR-AUC | 0.9574 |
| ROC-AUC | 0.9897 |
| Precision | 0.8533 |
| Recall | 0.9552 |
| 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
## Dataset Size (before turn expansion)
Original rows (after cleaning and balancing): 1630 (unsafe: 340, safe: 1290)