Jailbreak Attack on StrongREJECT (test)
82.34ScoreGCD
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GCDVictim Model=MISTRAL-7B, Perplexity Defense=Disabled2026.06 | 82.34 | — | — | — | — | |
| GCDVictim Model=MISTRAL-7B, Perplexity Defense=Enabled2026.06 | 82.34 | — | — | — | — | |
| AUTODANVictim Model=MISTRAL-7B, Perplexity Defense=Disabled2026.06 | 81.4 | — | — | — | — | |
| AUTODANVictim Model=MISTRAL-7B, Perplexity Defense=Enabled2026.06 | 81.4 | — | — | — | — | |
| AUTODANVictim Model=QWEN-2.5-7B, Perplexity Defense=Disabled2026.06 | 80.54 | — | — | — | — | |
| AUTODANVictim Model=QWEN-2.5-7B, Perplexity Defense=Enabled2026.06 | 80.54 | — | — | — | — | |
| GCDVictim Model=QWEN-2.5-7B, Perplexity Defense=Disabled2026.06 | 77.89 | — | — | — | — | |
| GCDVictim Model=QWEN-2.5-7B, Perplexity Defense=Enabled2026.06 | 77.89 | — | — | — | — | |
| INPAINTINGVictim Model=MISTRAL-7B, Perplexity Defense=Disabled2026.06 | 73.81 | — | — | — | — | |
| INPAINTINGVictim Model=MISTRAL-7B, Perplexity Defense=Enabled2026.06 | 73.81 | — | — | — | — | |
| GCDVictim Model=LLAMA-3-8B, Perplexity Defense=Disabled2026.06 | 71.59 | — | — | — | — | |
| GCDVictim Model=LLAMA-3-8B, Perplexity Defense=Enabled2026.06 | 71.59 | — | — | — | — | |
| GCDVictim Model=LLAMA-3-70B, Perplexity Defense=Disabled2026.06 | 65.71 | — | — | — | — | |
| GCDVictim Model=LLAMA-3-70B, Perplexity Defense=Enabled2026.06 | 65.71 | — | — | — | — | |
| GCGVictim Model=MISTRAL-7B, Perplexity Defense=Disabled2026.06 | 64.4 | — | — | — | — | |
| GBDAVictim Model=MISTRAL-7B, Perplexity Defense=Disabled2026.06 | 56.3 | — | — | — | — | |
| Direct RequestVictim Model=MISTRAL-7B, Perplexity Defense=Disabled2026.06 | 53.2 | — | — | — | — | |
| PAIRVictim Model=QWEN-2.5-7B, Perplexity Defense=Disabled2026.06 | 51.57 | — | — | — | — | |
| PAIRVictim Model=QWEN-2.5-7B, Perplexity Defense=Enabled2026.06 | 51.57 | — | — | — | — | |
| TAPVictim Model=QWEN-2.5-7B, Perplexity Defense=Disabled2026.06 | 50.84 | — | — | — | — | |
| TAPVictim Model=QWEN-2.5-7B, Perplexity Defense=Enabled2026.06 | 50.84 | — | — | — | — | |
| Direct RequestVictim Model=MISTRAL-7B, Perplexity Defense=Enabled2026.06 | 50.82 | — | — | — | — | |
| PAIRVictim Model=MISTRAL-7B, Perplexity Defense=Disabled2026.06 | 50.05 | — | — | — | — | |
| PAIRVictim Model=MISTRAL-7B, Perplexity Defense=Enabled2026.06 | 50.05 | — | — | — | — | |
| TAPVictim Model=MISTRAL-7B, Perplexity Defense=Disabled2026.06 | 49.09 | — | — | — | — | |
| TAPVictim Model=MISTRAL-7B, Perplexity Defense=Enabled2026.06 | 49.09 | — | — | — | — | |
| GCGVictim Model=QWEN-2.5-7B, Perplexity Defense=Disabled2026.06 | 48.54 | — | — | — | — | |
| INPAINTINGVictim Model=QWEN-2.5-7B, Perplexity Defense=Disabled2026.06 | 43.06 | — | — | — | — | |
| INPAINTINGVictim Model=QWEN-2.5-7B, Perplexity Defense=Enabled2026.06 | 43.06 | — | — | — | — | |
| PAIRVictim Model=LLAMA-3-70B, Perplexity Defense=Disabled2026.06 | 30.33 | — | — | — | — | |
| PAIRVictim Model=LLAMA-3-70B, Perplexity Defense=Enabled2026.06 | 30.33 | — | — | — | — | |
| INPAINTINGVictim Model=LLAMA-3-70B, Perplexity Defense=Disabled2026.06 | 26.38 | — | — | — | — | |
| INPAINTINGVictim Model=LLAMA-3-70B, Perplexity Defense=Enabled2026.06 | 26.38 | — | — | — | — | |
| INPAINTINGVictim Model=LLAMA-3-8B, Perplexity Defense=Disabled2026.06 | 15.75 | — | — | — | — | |
| INPAINTINGVictim Model=LLAMA-3-8B, Perplexity Defense=Enabled2026.06 | 15.75 | — | — | — | — | |
| AUTODANVictim Model=LLAMA-3-8B, Perplexity Defense=Disabled2026.06 | 15.68 | — | — | — | — | |
| AUTODANVictim Model=LLAMA-3-8B, Perplexity Defense=Enabled2026.06 | 15.68 | — | — | — | — | |
| PAIRVictim Model=LLAMA-3-8B, Perplexity Defense=Disabled2026.06 | 14.36 | — | — | — | — | |
| PAIRVictim Model=LLAMA-3-8B, Perplexity Defense=Enabled2026.06 | 14.36 | — | — | — | — | |
| TAPVictim Model=LLAMA-3-70B, Perplexity Defense=Disabled2026.06 | 12.94 | — | — | — | — | |
| TAPVictim Model=LLAMA-3-70B, Perplexity Defense=Enabled2026.06 | 12.94 | — | — | — | — | |
| Direct RequestVictim Model=QWEN-2.5-7B, Perplexity Defense=Disabled2026.06 | 11.99 | — | — | — | — | |
| Direct RequestVictim Model=QWEN-2.5-7B, Perplexity Defense=Enabled2026.06 | 11.99 | — | — | — | — | |
| GBDAVictim Model=QWEN-2.5-7B, Perplexity Defense=Disabled2026.06 | 9.34 | — | — | — | — | |
| TAPVictim Model=LLAMA-3-8B, Perplexity Defense=Disabled2026.06 | 5.53 | — | — | — | — | |
| TAPVictim Model=LLAMA-3-8B, Perplexity Defense=Enabled2026.06 | 5.53 | — | — | — | — | |
| Direct RequestVictim Model=LLAMA-3-70B, Perplexity Defense=Disabled2026.06 | 5.33 | — | — | — | — | |
| Direct RequestVictim Model=LLAMA-3-70B, Perplexity Defense=Enabled2026.06 | 5.33 | — | — | — | — | |
| GCGVictim Model=LLAMA-3-70B, Perplexity Defense=Disabled2026.06 | 3.69 | — | — | — | — | |
| Direct RequestVictim Model=LLAMA-3-8B, Perplexity Defense=Disabled2026.06 | 3.68 | — | — | — | — | |
| Direct RequestVictim Model=LLAMA-3-8B, Perplexity Defense=Enabled2026.06 | 3.68 | — | — | — | — | |
| GBDAVictim Model=LLAMA-3-70B, Perplexity Defense=Disabled2026.06 | 3.54 | — | — | — | — | |
| GCGVictim Model=LLAMA-3-8B, Perplexity Defense=Disabled2026.06 | 1.04 | — | — | — | — | |
| GBDAVictim Model=LLAMA-3-8B, Perplexity Defense=Disabled2026.06 | 0.39 | — | — | — | — | |
| AUTODANVictim Model=LLAMA-3-70B, Perplexity Defense=Disabled2026.06 | 0.37 | — | — | — | — | |
| AUTODANVictim Model=LLAMA-3-70B, Perplexity Defense=Enabled2026.06 | 0.37 | — | — | — | — | |
| GCGVictim Model=LLAMA-3-8B, Perplexity Defense=Enabled2026.06 | 0 | — | — | — | — | |
| GBDAVictim Model=LLAMA-3-8B, Perplexity Defense=Enabled2026.06 | 0 | — | — | — | — | |
| GCGVictim Model=QWEN-2.5-7B, Perplexity Defense=Enabled2026.06 | 0 | — | — | — | — | |
| GBDAVictim Model=QWEN-2.5-7B, Perplexity Defense=Enabled2026.06 | 0 | — | — | — | — | |
| GCGVictim Model=MISTRAL-7B, Perplexity Defense=Enabled2026.06 | 0 | — | — | — | — | |
| GBDAVictim Model=MISTRAL-7B, Perplexity Defense=Enabled2026.06 | 0 | — | — | — | — | |
| GCGVictim Model=LLAMA-3-70B, Perplexity Defense=Enabled2026.06 | 0 | — | — | — | — | |
| GBDAVictim Model=LLAMA-3-70B, Perplexity Defense=Enabled2026.06 | 0 | — | — | — | — | |
| CKA-AgentAttack model=Qwen3-32B-abliterated (Thinking), Judge model=Gemini-2.5-Pro, Method type=Decomposition2025.12 | — | 98.8 | 0.6 | 0 | 0.6 | |
| CKA-AgentAttack model=Qwen3-32B-abliterated (Thinking), Judge model=GPT-5-mini, Method type=Decomposition2025.12 | — | 97.5 | 2.5 | 0 | 0 | |
| CKA-AgentAttack model=Qwen3-32B-abliterated (Thinking), Judge model=Fine-Tuned Judge (Gemma-2b), Method type=Decomposition2025.12 | — | 98.8 | 1.2 | 0 | 0 |