Question Answering on WIKI-BALANCE (test)
66Avg. PS Shift (%)BRRA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BRRALLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=12025.11 | 66 | — | |
| No Def.LLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=12025.11 | 67 | — | |
| SMARTLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=102025.11 | 68 | — | |
| SMARTLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=12025.11 | 79 | — | |
| SMARTLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=52025.11 | 80 | — | |
| BRRALLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=12025.11 | 83 | — | |
| MMRLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=102025.11 | 85 | — | |
| MMRLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=52025.11 | 87 | — | |
| MMRLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=12025.11 | 89 | — | |
| No Def.LLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=12025.11 | 92 | — | |
| SMARTLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=12025.11 | 98 | — | |
| SMARTLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=102025.11 | 100 | — | |
| SMARTLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=52025.11 | 105 | — | |
| MMRLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=52025.11 | 118 | — | |
| BRRALLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=52025.11 | 122 | — | |
| MMRLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=12025.11 | 124 | — | |
| No Def.LLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=52025.11 | 125 | — | |
| BRRALLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=102025.11 | 129 | — | |
| No Def.LLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Negative Injection, Number of Injected Passages=102025.11 | 132 | — | |
| MMRLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=102025.11 | 139 | — | |
| BRRALLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=52025.11 | 202 | — | |
| No Def.LLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=52025.11 | 203 | — | |
| No Def.LLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=102025.11 | 203 | — | |
| BRRALLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=Positive Injection, Number of Injected Passages=102025.11 | 212 | — | |
| BRRALLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=None2025.11 | — | 0.058 | |
| MMRLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=None2025.11 | — | 0.054 | |
| No Def.LLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=None2025.11 | — | 0.06 | |
| SMARTLLM Model=DeepSeek-R1-Distill-Qwen-14B, Attack Type=None2025.11 | — | 0.056 |