Jailbreak detection on GPF jailbreak prompts
97AccuracyFrameShield-Crit
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FrameShield-CritLLM Model=Qwen2-0.5B2026.02 | 97 | 94 | |
| FrameShield-LastLLM Model=Qwen2-0.5B2026.02 | 96 | 92 | |
| JBShieldLLM Model=Qwen3-4B2026.02 | 87 | 88 | |
| FrameShield-CritLLM Model=Qwen3-4B2026.02 | 85 | 71 | |
| FrameShield-LastLLM Model=Qwen3-4B2026.02 | 84 | 68 | |
| JBShieldLLM Model=Qwen2-0.5B2026.02 | 83 | 85 | |
| JBShieldLLM Model=Qwen2.5-7B2026.02 | 83 | 85 | |
| FrameShield-CritLLM Model=Qwen2.5-7B2026.02 | 83 | 64 | |
| FrameShield-LastLLM Model=Qwen2.5-7B2026.02 | 81 | 73 |