Jailbreak Attack on Qwen3Guard Gen 4B
96ASRBlock 1 MLP Probe
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Block 1 MLP ProbeBlock=1, Probe Type=MLP2026.05 | 96 | 0.32 | |
| Block 1 LR ProbeBlock=1, Probe Type=LR2026.05 | 93 | 0.32 | |
| Block 2 LR ProbeBlock=2, Probe Type=LR2026.05 | 92 | 0.31 | |
| Block 2 MLP ProbeBlock=2, Probe Type=MLP2026.05 | 91 | 0.32 | |
| Block 10 MLP ProbeBlock=10, Probe Type=MLP2026.05 | 89 | 0.43 | |
| Vanilla AutoDANAttack Type=Vanilla AutoDAN2026.05 | 87 | 0.8 | |
| Block 17 LR ProbeBlock=17, Probe Type=LR2026.05 | 87 | 0.49 | |
| Block 10 LR ProbeBlock=10, Probe Type=LR2026.05 | 84 | 0.48 | |
| Block 17 MLP ProbeBlock=17, Probe Type=MLP2026.05 | 82 | 0.5 | |
| Template OnlyAttack Type=Template Only2026.05 | 4 | — |