Indirect Prompt Injection on Multi-News
100ASRJudgeDeceiver
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| JudgeDeceiverTarget LLM=Qwen3-4B2025.12 | 100 | — | |
| ObliInjection-GCGTarget LLM=Llama4-17B2025.12 | 100 | — | |
| ObliInjectionTarget LLM=Llama-3-8B2025.12 | 100 | — | |
| ObliInjectionTarget LLM=Llama4-17B2025.12 | 100 | — | |
| ObliInjectionTarget LLM=Qwen3-4B2025.12 | 100 | — | |
| ObliInjection-CETarget LLM=Falcon3-7B2025.12 | 99.8 | — | |
| ObliInjection-GCGTarget LLM=Qwen3-4B2025.12 | 99.6 | — | |
| ObliInjection-CETarget LLM=Llama-3-8B2025.12 | 99.6 | — | |
| ObliInjection-CETarget LLM=Llama4-17B2025.12 | 99.6 | — | |
| ObliInjectionTarget LLM=Llama-3.1-8B2025.12 | 99.6 | — | |
| ObliInjectionTarget LLM=Qwen-2.5-7B2025.12 | 99.6 | — | |
| ObliInjection-CETarget LLM=Llama-3.1-8B2025.12 | 99.4 | — | |
| ObliInjectionTarget LLM=Falcon3-7B2025.12 | 98 | — | |
| Neural ExecTarget LLM=Llama4-17B2025.12 | 97 | — | |
| ObliInjection-GCGTarget LLM=Qwen-2.5-7B2025.12 | 94.2 | — | |
| JudgeDeceiverTarget LLM=Qwen-2.5-7B2025.12 | 94 | — | |
| ObliInjectionTarget LLM=Mistral-7B2025.12 | 93.8 | — | |
| ObliInjection-CETarget LLM=Qwen3-4B2025.12 | 91 | — | |
| Neural ExecTarget LLM=Qwen3-4B2025.12 | 89 | — | |
| Combined AttackTarget LLM=Qwen3-4B2025.12 | 87.5 | — | |
| ObliInjection-GCGTarget LLM=Llama-3-8B2025.12 | 82 | — | |
| JudgeDeceiverTarget LLM=Llama-3-8B2025.12 | 76.6 | — | |
| ObliInjection-CETarget LLM=Mistral-7B2025.12 | 67 | — | |
| Neural ExecTarget LLM=Llama-3-8B2025.12 | 58.4 | — | |
| Combined AttackTarget LLM=Llama4-17B2025.12 | 51.6 | — | |
| ObliInjection-CETarget LLM=Qwen-2.5-7B2025.12 | 40.4 | — | |
| ObliInjection-GCGTarget LLM=Llama-3.1-8B2025.12 | 34.8 | — | |
| ObliInjection-GCGTarget LLM=Falcon3-7B2025.12 | 19.6 | — | |
| Combined AttackTarget LLM=Falcon3-7B2025.12 | 8.4 | — | |
| Neural ExecTarget LLM=Mistral-7B2025.12 | 7.8 | — | |
| Combined AttackTarget LLM=Qwen-2.5-7B2025.12 | 7 | — | |
| JudgeDeceiverTarget LLM=Falcon3-7B2025.12 | 5 | — | |
| Neural ExecTarget LLM=Llama-3.1-8B2025.12 | 3 | — | |
| JudgeDeceiverTarget LLM=Mistral-7B2025.12 | 2.6 | — | |
| JudgeDeceiverTarget LLM=Llama4-17B2025.12 | 2.4 | — | |
| Combined AttackTarget LLM=Llama-3.1-8B2025.12 | 0.6 | — | |
| Neural ExecTarget LLM=Falcon3-7B2025.12 | 0.4 | — | |
| Combined AttackTarget LLM=Llama-3-8B2025.12 | 0 | — | |
| Combined AttackTarget LLM=Mistral-7B2025.12 | 0 | — | |
| Neural ExecTarget LLM=Qwen-2.5-7B2025.12 | 0 | — | |
| JudgeDeceiverTarget LLM=Llama-3.1-8B2025.12 | 0 | — | |
| ObliInjection-GCGTarget LLM=Mistral-7B2025.12 | 0 | — | |
| Combined AttackTarget LLM=Average2025.12 | — | 22.2 | |
| JudgeDeceiverTarget LLM=Average2025.12 | — | 40.1 | |
| Neural ExecTarget LLM=Average2025.12 | — | 36.5 | |
| ObliInjectionTarget LLM=Average2025.12 | — | 98.7 | |
| ObliInjection-CETarget LLM=Average2025.12 | — | 85.3 | |
| ObliInjection-GCGTarget LLM=Average2025.12 | — | 61.5 |