Jailbreak Detection on ALL-4
92.6Accuracygpt-5-mini
Evaluation Results
| Method | Links | |
|---|---|---|
| gpt-5-miniVersion=2025-08-072026.02 | 92.6 | |
| AISABackbone=Llama3.1-8b-I2026.02 | 88.89 | |
| AISABackbone=Llama2-13b-I2026.02 | 87.36 | |
| gpt-4o-miniVersion=2024-07-182026.02 | 85.89 | |
| AISABackbone=Qwen3-8b-I2026.02 | 85.82 | |
| Jailbreak-Classifier2026.02 | 85 | |
| AISABackbone=GPT-OSS-20b-I2026.02 | 83.95 | |
| AISABackbone=Mistral-7b-I2026.02 | 82.65 | |
| gpt-4.1-miniVersion=2025-04-142026.02 | 81.05 | |
| NemoGuard-JailbreakDetect2026.02 | 72.9 | |
| GradSafe2026.02 | 69.8 | |
| SPDetector2026.02 | 54.64 | |
| Llama-Prompt-Guard-2Parameters=86M2026.02 | 54.43 |