Jailbreak Evaluation on StrongReject
95.5ASR-Jτ_trigger ⊕ PAP
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| τ_trigger ⊕ PAPModel=Backdoor Model (ACB), Backdoor Trigger=τ_trigger2026.04 | 95.5 | 83.7 | 74.5 | — | |
| τ_trigger ⊕ TAPModel=Backdoor Model (ACB), Backdoor Trigger=τ_trigger2026.04 | 91.7 | 81.7 | 60.4 | — | |
| τ_trigger ⊕ PAIRModel=Backdoor Model (ACB), Backdoor Trigger=τ_trigger2026.04 | 88.1 | 82.1 | 59.2 | — | |
| PAPModel=Qwen2.5-7B-Instruct, Backdoor Trigger=None2026.04 | 84.2 | 77 | 64.8 | — | |
| τ_trigger ⊕ DirectModel=Backdoor Model (ACB), Backdoor Trigger=τ_trigger2026.04 | 72 | 49.2 | 39.5 | — | |
| TAPModel=Qwen2.5-7B-Instruct, Backdoor Trigger=None2026.04 | 66.5 | 62.6 | 51.4 | — | |
| PAIRModel=Qwen2.5-7B-Instruct, Backdoor Trigger=None2026.04 | 61.9 | 59 | 49.7 | — | |
| τ_trigger ⊕ DirectModel=Qwen2.5-7B-Instruct, Backdoor Trigger=τ_trigger2026.04 | 21.4 | 13.7 | 8.8 | — | |
| DirectModel=Qwen2.5-7B-Instruct, Backdoor Trigger=None2026.04 | 16.9 | 11.5 | 9.2 | — | |
| DAPO2025.10 | — | — | — | 3.2 | |
| RECAP2025.10 | — | — | — | 0.3 | |
| SAFE-LLAMA-8BBase Model=Llama-3-8B2025.10 | — | — | — | 8.1 | |
| SafeChain2025.10 | — | — | — | 22.3 | |
| STAR2025.10 | — | — | — | 7.8 |