RAG Attack on MS-MARCO
0.8Attack Success Rate (ASR)MuteRAG
Evaluation Results
| Method | Links | |
|---|---|---|
| MuteRAGAttack Type=Blocking, Target LLM=GPT-5-mini2026.03 | 0.8 | |
| MuteRAGAttack Type=Blocking, Target LLM=GPT-5.22026.03 | 0.9 | |
| MuteRAGAttack Type=Blocking, Target LLM=Ministral-3-8B2026.03 | 3 | |
| MuteRAGAttack Type=Blocking, Target LLM=DeepSeek-V3.22026.03 | 3.8 | |
| MuteRAGAttack Type=Blocking, Target LLM=Qwen-3-32B2026.03 | 4.4 | |
| AuthChainAttack Type=Misleading, Target LLM=GPT-5-mini2026.03 | 10 | |
| AuthChainAttack Type=Misleading, Target LLM=GPT-5.22026.03 | 11.3 | |
| MuteRAGAttack Type=Blocking, Target LLM=Gemma-3-12B2026.03 | 11.3 | |
| AuthChainAttack Type=Misleading, Target LLM=DeepSeek-V3.22026.03 | 13.3 | |
| AuthChainAttack Type=Misleading, Target LLM=Qwen-3-32B2026.03 | 14.7 | |
| AuthChainAttack Type=Misleading, Target LLM=Ministral-3-8B2026.03 | 15.3 | |
| AuthChainAttack Type=Misleading, Target LLM=Gemma-3-12B2026.03 | 16 | |
| AuthChainAttack Type=Misleading, Target LLM=Llama-3-8B2026.03 | 18 | |
| JammingAttack Type=Blocking, Target LLM=GPT-5.22026.03 | 19.4 | |
| JammingAttack Type=Blocking, Target LLM=Qwen-3-32B2026.03 | 23.1 | |
| JammingAttack Type=Blocking, Target LLM=DeepSeek-V3.22026.03 | 24.4 | |
| JammingAttack Type=Blocking, Target LLM=GPT-5-mini2026.03 | 25 | |
| JammingAttack Type=Blocking, Target LLM=Gemma-3-12B2026.03 | 29.3 | |
| JammingAttack Type=Blocking, Target LLM=Llama-3-8B2026.03 | 29.6 | |
| TabooRAGAttack Type=Blocking, Target LLM=Qwen-3-32B2026.03 | 44.5 | |
| PoisonedAttack Type=Misleading, Target LLM=Gemma-3-12B2026.03 | 46.7 | |
| JammingAttack Type=Blocking, Target LLM=Ministral-3-8B2026.03 | 46.7 | |
| PoisonedAttack Type=Misleading, Target LLM=GPT-5.22026.03 | 49.3 | |
| PoisonedAttack Type=Misleading, Target LLM=Ministral-3-8B2026.03 | 52 | |
| PoisonedAttack Type=Misleading, Target LLM=GPT-5-mini2026.03 | 52 | |
| PoisonedAttack Type=Misleading, Target LLM=DeepSeek-V3.22026.03 | 52.7 | |
| TabooRAGAttack Type=Blocking, Target LLM=Ministral-3-8B2026.03 | 53.8 | |
| TabooRAGAttack Type=Blocking, Target LLM=Gemma-3-12B2026.03 | 54 | |
| PoisonedAttack Type=Misleading, Target LLM=Llama-3-8B2026.03 | 55.3 | |
| TabooRAGAttack Type=Blocking, Target LLM=Llama-3-8B2026.03 | 55.6 | |
| PoisonedAttack Type=Misleading, Target LLM=Qwen-3-32B2026.03 | 56 | |
| TabooRAGAttack Type=Blocking, Target LLM=GPT-5-mini2026.03 | 56.3 | |
| TabooRAGAttack Type=Blocking, Target LLM=DeepSeek-V3.22026.03 | 64.6 | |
| TabooRAGAttack Type=Blocking, Target LLM=GPT-5.22026.03 | 68.9 | |
| MuteRAGAttack Type=Blocking, Target LLM=Llama-3-8B2026.03 | 74.1 |