Jailbreak Detection on JBC
99.2AccuracyLlama-Prompt-Guard-2
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-Prompt-Guard-2Parameters=86M2026.02 | 99.2 | |
| AISABackbone=Qwen3-8b-I2026.02 | 97.99 | |
| Jailbreak-Classifier2026.02 | 97.19 | |
| AISABackbone=Llama3.1-8b-I2026.02 | 96.39 | |
| AISABackbone=GPT-OSS-20b-I2026.02 | 96.39 | |
| AISABackbone=Mistral-7b-I2026.02 | 95.98 | |
| AISABackbone=Llama2-13b-I2026.02 | 95.58 | |
| gpt-4.1-miniVersion=2025-04-142026.02 | 91.57 | |
| gpt-5-miniVersion=2025-08-072026.02 | 89.16 | |
| NemoGuard-JailbreakDetect2026.02 | 87.55 | |
| gpt-4o-miniVersion=2024-07-182026.02 | 87.15 | |
| GradSafe2026.02 | 64.66 | |
| SPDetector2026.02 | 55.02 |