Jailbreak Attack on Qwen 27B 3.6
89ASRVanilla AutoDAN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Vanilla AutoDANAttack Type=Gradient-based search2026.05 | 89 | 17.44 | |
| Block 10 LR ProbeMethod=Mechanistic AutoDAN, Probe Type=Logistic Regression, Block=102026.05 | 82 | 4.79 | |
| Block 29 MLP ProbeMethod=Mechanistic AutoDAN, Probe Type=Multi-Layer Perceptron, Block=292026.05 | 82 | 7.05 | |
| Block 10 MLP ProbeMethod=Mechanistic AutoDAN, Probe Type=Multi-Layer Perceptron, Block=102026.05 | 79 | 7.05 | |
| Block 32 LR ProbeMethod=Mechanistic AutoDAN, Probe Type=Logistic Regression, Block=322026.05 | 73 | 8.18 | |
| Block 1 LR ProbeMethod=Mechanistic AutoDAN, Probe Type=Logistic Regression, Block=12026.05 | 32 | 1.77 | |
| Block 1 MLP ProbeMethod=Mechanistic AutoDAN, Probe Type=Multi-Layer Perceptron, Block=12026.05 | 28 | 1.47 | |
| Template OnlyAttack Type=Baseline2026.05 | 0 | — |