Question Classification on TREC (Adversarial Metrics)
0.938Attack Success Rate (ASR)CacheTrap
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CacheTrapModel=Mistral-7B, Target Class=12025.11 | 0.938 | — | |
| CacheTrapModel=Mistral-7B, Target Class=32025.11 | 0.938 | — | |
| CacheTrapModel=Mistral-7B, Target Class=22025.11 | 0.968 | — | |
| CacheTrapModel=Mistral-7B, Target Class=02025.11 | 0.986 | 0.97 | |
| CacheTrapModel=LLaMA-2-7B, Target Class=02025.11 | 1 | 0.97 | |
| CacheTrapModel=LLaMA-2-7B, Target Class=12025.11 | 1 | — | |
| CacheTrapModel=LLaMA-2-7B, Target Class=22025.11 | 1 | — | |
| CacheTrapModel=LLaMA-2-7B, Target Class=32025.11 | 1 | — | |
| CacheTrapModel=LLaMA-2-7B, Target Class=42025.11 | 1 | — | |
| CacheTrapModel=LLaMA-2-7B, Target Class=52025.11 | 1 | — | |
| CacheTrapModel=LLaMA-3.1-8B, Target Class=02025.11 | 1 | 0.972 | |
| CacheTrapModel=Qwen-2.5-3B, Target Class=02025.11 | 1 | 0.966 | |
| CacheTrapModel=DeepSeek-7B, Target Class=02025.11 | 1 | 0.948 |