Defense Robustness on GCG Attack
100Keyword Match RateES2
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ES2LLM=*Llama-2 (7B-Chat)2026.03 | 100 | 100 | 92 | |
| ES2LLM=Llama-3 (8B-Instruct)2026.03 | 100 | 100 | 98 | |
| DPLLLM=*Llama-2 (7B-Chat)2026.03 | 99 | 99 | 88 | |
| ES2LLM=Qwen-2.5 (7B-Instruct)2026.03 | 98 | 98 | 94 | |
| STLLLM=Llama-3 (8B-Instruct)2026.03 | 97 | 95 | 93 | |
| Base ModelLLM=Llama-3 (8B-Instruct)2026.03 | 96 | 94 | 84 | |
| DPLLLM=Llama-3 (8B-Instruct)2026.03 | 96 | 94 | 93 | |
| STLLLM=*Llama-2 (7B-Chat)2026.03 | 90 | 88 | 73 | |
| Base ModelLLM=*Llama-2 (7B-Chat)2026.03 | 89 | 88 | 78 | |
| DPLLLM=Qwen-2.5 (7B-Instruct)2026.03 | 89 | 88 | 84 | |
| STLLLM=Qwen-2.5 (7B-Instruct)2026.03 | 88 | 87 | 87 | |
| Base ModelLLM=Qwen-2.5 (7B-Instruct)2026.03 | 79 | 74 | 73 | |
| ES2LLM=Mistral (7B-Instruct)2026.03 | 56 | 50 | 44 | |
| DPLLLM=Mistral (7B-Instruct)2026.03 | 28 | 24 | 22 | |
| STLLLM=Mistral (7B-Instruct)2026.03 | 25 | 24 | 24 | |
| Base ModelLLM=Mistral (7B-Instruct)2026.03 | 21 | 19 | 19 |