Jailbreak Attack on Targeted Model Jailbreak Evaluation Suite v1.0
81.6ASRCALM
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| CALMTarget Model=Qwen-Turbo, Method Category=Topic-Free2025.06 | 81.6 | 21.65 | 47.71 | 1.28 | 17.67 | 38.93 | 1.05 | |
| CRTTarget Model=Qwen-Turbo, Method Category=Topic-Free2025.06 | 80 | 20.71 | 42.64 | 1.28 | 16.57 | 34.12 | 1.02 | |
| DiveR-CTTarget Model=Qwen-Turbo, Method Category=Topic-Free2025.06 | 79.6 | 22.82 | 46.29 | 1.28 | 18.17 | 36.85 | 1.02 | |
| RedTopicTarget Model=r1-Qwen-14b, Method Category=Topic-Free2025.06 | 62.05 | 23.13 | 63.14 | 6.57 | 14.35 | 39.18 | 4.08 | |
| RedTopicTarget Model=LLaMA-3b, Method Category=Topic-Free2025.06 | 60.85 | 23.48 | 62.64 | 7.8 | 14.29 | 38.12 | 4.75 | |
| RedTopicTarget Model=Gemma-9b, Method Category=Topic-Free2025.06 | 47.65 | 21.87 | 62.07 | 9.23 | 10.42 | 29.57 | 4.4 | |
| RedTopicTarget Model=deepseek-r1, Method Category=Topic-Free2025.06 | 42.6 | 21.18 | 48.94 | 8.48 | 9.02 | 20.85 | 3.67 | |
| Latent-JailbreakTarget Model=Qwen-Turbo, Method Category=Topic-Based2025.06 | 34.8 | 0.1 | 23.74 | 1.62 | 0.03 | 8.26 | 0.56 | |
| RedTopicTarget Model=Qwen-Turbo, Method Category=Topic-Free2025.06 | 29.5 | 23.22 | 62.54 | 11 | 6.85 | 18.45 | 3.25 | |
| RedTopicTarget Model=Gemini-2.0, Method Category=Topic-Free2025.06 | 24.6 | 24.65 | 65.38 | 12.13 | 6.06 | 16.08 | 2.98 | |
| RedTopicTarget Model=GPT-4o, Method Category=Topic-Free2025.06 | 23.6 | 21.91 | 51.62 | 13.89 | 5.17 | 12.18 | 3.23 | |
| RFTTarget Model=Qwen-Turbo, Method Category=Topic-Free2025.06 | 20.7 | 2.7 | 5.94 | 4.42 | 0.56 | 1.23 | 0.91 | |
| Wild-Jailbreak (wild-2)Target Model=Qwen-Turbo, Method Category=Topic-Based2025.06 | 10.15 | 0 | 5.14 | 6.62 | 0 | 0.77 | 0.99 | |
| Wild-Jailbreak (wild-1)Target Model=Qwen-Turbo, Method Category=Topic-Based2025.06 | 9.5 | 0 | 7.69 | 8.02 | 0 | 1.15 | 1.2 | |
| ASTRALTarget Model=Qwen-Turbo, Method Category=Topic-Based2025.06 | 6.5 | 0 | 55.33 | 14.91 | 0 | 8.3 | 2.24 | |
| JailbreakV-28KTarget Model=Qwen-Turbo, Method Category=Topic-Based2025.06 | 3.45 | 0 | 42.35 | 10.66 | 0 | 6.35 | 1.6 |