Adversarial Attack on AdvBench Original (test)
0CHRFPFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FPFTModel=DeepSeek-7B2026.04 | 0 | 10 | |
| TSBH (Pass@1)Model=DeepSeek-7B2026.04 | 0 | 0 | |
| TSBH (Pass@5)Model=DeepSeek-7B2026.04 | 0 | 0 | |
| FPFTModel=Qwen2.5-7B2026.04 | 0 | 23 | |
| TSBH (Pass@1)Model=Qwen2.5-7B2026.04 | 0 | 3 | |
| TSBH (Pass@5)Model=Qwen2.5-7B2026.04 | 0 | 2 | |
| FPFTModel=Llama3.2-3B2026.04 | 0 | 38 | |
| TSBH (Pass@1)Model=Llama3.2-3B2026.04 | 0 | 0 | |
| TSBH (Pass@5)Model=Llama3.2-3B2026.04 | 0 | 0 | |
| H-CoTModel=Qwen2.5-7B2026.04 | 1 | 1 | |
| AutoRANModel=Qwen2.5-7B2026.04 | 1 | 1 | |
| Chain-of-LureModel=Qwen2.5-7B2026.04 | 1 | 1 | |
| CoT HijackingModel=Qwen2.5-7B2026.04 | 1 | 1 | |
| OriginalModel=Llama3.2-3B2026.04 | 1 | 1 | |
| H-CoTModel=Llama3.2-3B2026.04 | 1 | 1 | |
| AutoRANModel=Llama3.2-3B2026.04 | 1 | 1 | |
| Chain-of-LureModel=Llama3.2-3B2026.04 | 1 | 1 | |
| CoT HijackingModel=Llama3.2-3B2026.04 | 1 | 1 | |
| H-CoTModel=DeepSeek-7B2026.04 | 3 | 4 | |
| AutoRANModel=DeepSeek-7B2026.04 | 4 | 5 | |
| Chain-of-LureModel=DeepSeek-7B2026.04 | 4 | 4 | |
| CoT HijackingModel=DeepSeek-7B2026.04 | 4 | 4 | |
| LRFTModel=Qwen2.5-7B2026.04 | 16 | 14 | |
| OriginalModel=Qwen2.5-7B2026.04 | 23 | 7 | |
| OriginalModel=DeepSeek-7B2026.04 | 36 | 22 | |
| LRFTModel=DeepSeek-7B2026.04 | 49 | 54 | |
| LRFTModel=Llama3.2-3B2026.04 | 50 | 53 |