Jailbreak Attack on RedTeam 2K
94.79ASRMAPA
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MAPATarget Model=Qwen2.5-VL-7B-Instruct2026.02 | 94.79 | — | — | — | — | |
| MAPATarget Model=Llava V1.6-Mistral-7B2026.02 | 93.75 | — | — | — | — | |
| MAPATarget Model=Average2026.02 | 92.21 | — | — | — | — | |
| MAPATarget Model=Llama-3.2-Vision-11B-Instruct2026.02 | 88.09 | — | — | — | — | |
| CoATarget Model=Llava V1.6-Mistral-7B2026.02 | 68.33 | — | — | — | — | |
| CoATarget Model=Qwen2.5-VL-7B-Instruct2026.02 | 65.48 | — | — | — | — | |
| CoATarget Model=Average2026.02 | 65.24 | — | — | — | — | |
| FootInTheDoorTarget Model=Llama-3.2-Vision-11B-Instruct2026.02 | 63.33 | — | — | — | — | |
| CoATarget Model=Llama-3.2-Vision-11B-Instruct2026.02 | 61.9 | — | — | — | — | |
| FootInTheDoorTarget Model=Llava V1.6-Mistral-7B2026.02 | 55 | — | — | — | — | |
| FootInTheDoorTarget Model=Average2026.02 | 49.44 | — | — | — | — | |
| CMRMVLM=LLaVA-1.5-7B2026.03 | 47.5 | — | — | — | — | |
| ActorAttackTarget Model=Llama-3.2-Vision-11B-Instruct2026.02 | 47.37 | — | — | — | — | |
| CMRMVLM=LLaVA-1.5-7B2026.03 | 46.2 | — | — | — | — | |
| LLaVA-1.5-7B (Baseline)VLM=LLaVA-1.5-7B2026.03 | 39.3 | — | — | — | — | |
| LLaVA-1.5-7B (Baseline)VLM=LLaVA-1.5-7B2026.03 | 37.6 | — | — | — | — | |
| ShiftDCVLM=LLaVA-1.5-7B2026.03 | 33 | — | — | — | — | |
| ShiftDCVLM=LLaVA-1.5-7B2026.03 | 32.5 | — | — | — | — | |
| InternVL-Chat-19B (Baseline)VLM=InternVL-Chat-19B2026.03 | 32.2 | — | — | — | — | |
| InternVL-Chat-19B (Baseline)VLM=InternVL-Chat-19B2026.03 | 30.2 | — | — | — | — | |
| FootInTheDoorTarget Model=Qwen2.5-VL-7B-Instruct2026.02 | 30 | — | — | — | — | |
| CMRMVLM=InternVL-Chat-19B2026.03 | 29.8 | — | — | — | — | |
| ShareGPT4V-7B (Baseline)VLM=ShareGPT4V-7B2026.03 | 29.4 | — | — | — | — | |
| ShareGPT4V-7B (Baseline)VLM=ShareGPT4V-7B2026.03 | 28.8 | — | — | — | — | |
| ActorAttackTarget Model=Average2026.02 | 27.92 | — | — | — | — | |
| ActorAttackTarget Model=Qwen2.5-VL-7B-Instruct2026.02 | 25.86 | — | — | — | — | |
| ShiftDCVLM=InternVL-Chat-19B2026.03 | 25.5 | — | — | — | — | |
| AdaShieldVLM=InternVL-Chat-19B2026.03 | 24.7 | — | — | — | — | |
| AdaShieldVLM=LLaVA-1.5-7B2026.03 | 23.8 | — | — | — | — | |
| ECSOVLM=LLaVA-1.5-7B2026.03 | 23.2 | — | — | — | — | |
| ECSOVLM=ShareGPT4V-7B2026.03 | 22.7 | — | — | — | — | |
| ECSOVLM=LLaVA-1.5-7B2026.03 | 22.2 | — | — | — | — | |
| ECSOVLM=ShareGPT4V-7B2026.03 | 21.5 | — | — | — | — | |
| AdaShieldVLM=InternVL-Chat-19B2026.03 | 20.7 | — | — | — | — | |
| ShiftDCVLM=InternVL-Chat-19B2026.03 | 20.4 | — | — | — | — | |
| AdaShieldVLM=LLaVA-1.5-7B2026.03 | 19.7 | — | — | — | — | |
| JRS-RemVLM=LLaVA-1.5-7B2026.03 | 19 | — | — | — | — | |
| JRS-RemVLM=LLaVA-1.5-7B2026.03 | 18.4 | — | — | — | — | |
| ShiftDCVLM=ShareGPT4V-7B2026.03 | 17.9 | — | — | — | — | |
| CMRMVLM=InternVL-Chat-19B2026.03 | 17.8 | — | — | — | — | |
| ShiftDCVLM=ShareGPT4V-7B2026.03 | 17.6 | — | — | — | — | |
| ECSOVLM=InternVL-Chat-19B2026.03 | 14.4 | — | — | — | — | |
| JRS-RemVLM=InternVL-Chat-19B2026.03 | 14.3 | — | — | — | — | |
| CMRMVLM=ShareGPT4V-7B2026.03 | 12.5 | — | — | — | — | |
| AdaShieldVLM=ShareGPT4V-7B2026.03 | 12.3 | — | — | — | — | |
| AdaShieldVLM=ShareGPT4V-7B2026.03 | 12.2 | — | — | — | — | |
| CMRMVLM=ShareGPT4V-7B2026.03 | 12.1 | — | — | — | — | |
| ECSOVLM=InternVL-Chat-19B2026.03 | 11.2 | — | — | — | — | |
| ActorAttackTarget Model=Llava V1.6-Mistral-7B2026.02 | 10.53 | — | — | — | — | |
| JRS-RemVLM=InternVL-Chat-19B2026.03 | 8.7 | — | — | — | — | |
| JRS-RemVLM=ShareGPT4V-7B2026.03 | 8.5 | — | — | — | — | |
| JRS-RemVLM=ShareGPT4V-7B2026.03 | 8 | — | — | — | — | |
| MousetrapModel=Claude-3-5-Sonnet, Total Samples=20002025.02 | — | 43.85 | 91.7 | 98.1 | 1.9 |