List your company

Indirect Prompt Injection Attack on IPI-3k

69.3ASRAdapTools
-2.77215.93934.6553.361Feb 24, 2026

Evaluation results

MethodLinks
AdapToolsFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=No Defense2026.0269.36.7
AdapToolsFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=No Defense2026.0260.632.6
AutoHijackerFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=No Defense2026.0247.412.4
AdapToolsFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=No Defense2026.0244.415.2
AdapToolsFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=Pi-Detector2026.0241.324.3
Ignore InstructionFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=No Defense2026.0239.412.8
AdapToolsFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=Pi-Detector2026.0239.150
AutoHijackerFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=No Defense2026.0237.617.8
Combined AttackFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=No Defense2026.0237.414.2
AdapToolsFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=MELON2026.0236.427.3
AdapToolsFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=No Defense2026.0235.417.3
InjectAgentFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=No Defense2026.0235.214.4
InjectAgentFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=No Defense2026.0234.816
AdapToolsFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=MELON2026.0233.754.3
Combined AttackFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=No Defense2026.0232.817.8
AutoHijackerFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=No Defense2026.0229.649.6
AutoHijackerFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=Pi-Detector2026.0227.619.5
AdapToolsFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=Pi-Detector2026.0226.719.4
AdapToolsFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=No Defense2026.0226.144.8
Ignore InstructionFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=No Defense2026.0225.223
AutoHijackerFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=Pi-Detector2026.0223.820.3
AdapToolsFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=Pi-Detector2026.0223.724.4
Ignore InstructionFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=Pi-Detector2026.0223.218.6
InjectAgentFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=Pi-Detector2026.022319
AdapToolsFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=MELON2026.022320.9
Combined AttackFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=Pi-Detector2026.0222.619.5
AutoHijackerFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=No Defense2026.0222.426
InjectAgentFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=Pi-Detector2026.0220.621.1
Combined AttackFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=Pi-Detector2026.0220.422.5
AdapToolsFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=No Defense2026.0220.336.8
Combined AttackFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=No Defense2026.0219.456.4
AdapToolsFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=MELON2026.0218.527.6
Combined AttackFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=No Defense2026.0218.228.2
AutoHijackerFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=Pi-Detector2026.0216.460.2
Ignore InstructionFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=Pi-Detector2026.0216.226.4
AdapToolsFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=Pi-Detector2026.0216.151.1
InjectAgentFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=No Defense2026.021560.6
AutoHijackerFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=MELON2026.0214.824.1
Ignore InstructionFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=No Defense2026.0213.661
AdapToolsFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=Pi-Detector2026.0213.540.2
AutoHijackerFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=Pi-Detector2026.0213.431.5
AdapToolsFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=MELON2026.0213.352.9
AutoHijackerFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=MELON2026.0213.224.6
AutoHijackerFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=No Defense2026.021255.4
Ignore InstructionFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=MELON2026.0211.422.9
Combined AttackFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=MELON2026.0211.423.6
Combined AttackFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=Pi-Detector2026.021163.1
InjectAgentFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=MELON2026.0210.823.7
Combined AttackFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=Pi-Detector2026.021033.2
InjectAgentFM=MISTRAL-8B, CoT=None-reasoning, BU=36, Defense=MELON2026.029.625.1
Combined AttackFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=MELON2026.029.426.7
AutoHijackerFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=No Defense2026.029.340.2
Ignore InstructionFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=No Defense2026.029.231.4
InjectAgentFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=No Defense2026.029.233.2
InjectAgentFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=Pi-Detector2026.028.266
Combined AttackFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=No Defense2026.02854.4
Ignore InstructionFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=Pi-Detector2026.02865.5
Ignore InstructionFM=LLAMA-3.1-8B, CoT=None-reasoning, BU=38, Defense=MELON2026.027.829.6
AutoHijackerFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=MELON2026.027.467.4
AutoHijackerFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=Pi-Detector2026.026.859
AdapToolsFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=MELON2026.026.743.6
AutoHijackerFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=Pi-Detector2026.026.540.8
AutoHijackerFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=MELON2026.026.235.9
Ignore InstructionFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=Pi-Detector2026.025.233.8
Combined AttackFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=MELON2026.02536.3
InjectAgentFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=Pi-Detector2026.024.835.9
Combined AttackFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=MELON2026.024.868.1
Combined AttackFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=Pi-Detector2026.024.656.6
Ignore InstructionFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=MELON2026.024.468.4
InjectAgentFM=QWEN-3-8B, CoT=Reasoning, BU=80, Defense=MELON2026.024.469.1
AutoHijackerFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=MELON2026.024.341.3
InjectAgentFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=MELON2026.023.436.7
Ignore InstructionFM=GEMINI-2.5-FLASH, CoT=None-reasoning, BU=61, Defense=MELON2026.023.235.1
AutoHijackerFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=MELON2026.02361.5
Combined AttackFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=MELON2026.022.458.1
InjectAgentFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=No Defense2026.021.860.4
Ignore InstructionFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=No Defense2026.021.455.4
Ignore InstructionFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=Pi-Detector2026.020.855.8
InjectAgentFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=Pi-Detector2026.020.861.1
Ignore InstructionFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=MELON2026.020.456.1
InjectAgentFM=GPT-4.1, CoT=None-reasoning, BU=66, Defense=MELON2026.020.461.3
Ignore InstructionFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=No Defense2026.020.439.8
Combined AttackFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=No Defense2026.020.437.2
InjectAgentFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=No Defense2026.020.438
Ignore InstructionFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=Pi-Detector2026.020.239.9
Combined AttackFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=MELON2026.020.237.3
Combined AttackFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=Pi-Detector2026.020.237.3
InjectAgentFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=MELON2026.020.239.1
InjectAgentFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=Pi-Detector2026.020.238.1
Ignore InstructionFM=DEEPSEEK-R1, CoT=Reasoning, BU=50, Defense=MELON2026.02040