Jailbreaking on AdvBench Sub
4.73BERT ScoreNaive
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| NaiveModel=Llama3-8B-CB2025.05 | 4.73 | 1.85 | 0.65 | |
| LINTModel=Llama3-8B-CB2025.05 | 4.57 | 1.96 | 0.73 | |
| GCGModel=Llama3-8B-CB2025.05 | 4.36 | 2.19 | 0.5 | |
| FLIPModel=Llama3-8B-CB2025.05 | 4.3 | 2.08 | 0.42 | |
| JULIModel=Qwen2.5-1.5B2025.05 | 4.15 | 4.42 | 3.31 | |
| JULIModel=Llama3-8B-CB2025.05 | 4.08 | 2.38 | 2.04 | |
| EDModel=Llama3-8B-CB2025.05 | 3.99 | 3.73 | 2.54 | |
| BaseModel=Llama3-8B-CB2025.05 | 3.91 | 1.38 | 0.54 | |
| JULIModel=Llama3-3B-INST2025.05 | 3.9 | 4.27 | 2.85 | |
| Emulated AlignmentModel=Llama3-3B-INST2025.05 | 3.84 | 3.5 | 2.65 | |
| Emulated AlignmentModel=Llama3-8B-INST2025.05 | 3.84 | 3.81 | 2.42 | |
| Emulated AlignmentModel=Llama2-7B-CHAT2025.05 | 3.68 | 2.54 | 1.5 | |
| FLIPModel=Llama3-8B-INST2025.05 | 3.53 | 2.77 | 0.77 | |
| FLIPModel=Llama3-3B-INST2025.05 | 3.52 | 2.88 | 0.73 | |
| JULIModel=Llama2-7B-CHAT2025.05 | 3.35 | 3.85 | 2.38 | |
| GCGModel=Qwen2.5-1.5B2025.05 | 3.34 | 2.81 | 2 | |
| NaiveModel=Qwen2.5-1.5B2025.05 | 3.27 | 1.81 | 0.96 | |
| FLIPModel=Qwen2.5-1.5B2025.05 | 3.25 | 2.35 | 0.69 | |
| JULIModel=Llama3-8B-INST2025.05 | 3.12 | 3.62 | 2.45 | |
| Emulated AlignmentModel=Qwen2.5-1.5B2025.05 | 3.04 | 1.23 | 0.46 | |
| LINTModel=Qwen2.5-1.5B2025.05 | 3 | 3.69 | 1.65 | |
| WTSModel=Llama3-8B-CB2025.05 | 2.99 | 1.58 | 0.5 | |
| WTSModel=Qwen2.5-1.5B2025.05 | 2.89 | 3.15 | 1.73 | |
| BaseModel=Qwen2.5-1.5B2025.05 | 2.85 | 1.16 | 0.08 | |
| LINTModel=Llama3-8B-INST2025.05 | 2.7 | 3.5 | 1.84 | |
| BaseModel=Llama3-8B-INST2025.05 | 2.69 | 1.42 | 0.42 | |
| LINTModel=Llama2-7B-CHAT2025.05 | 2.67 | 3.5 | 1.35 | |
| NaiveModel=Llama3-8B-INST2025.05 | 2.46 | 1.81 | 0.85 | |
| NaiveModel=Llama2-7B-CHAT2025.05 | 2.42 | 2.81 | 2.19 | |
| GCGModel=Llama3-3B-INST2025.05 | 2.23 | 2.58 | 1.85 | |
| FLIPModel=Llama2-7B-CHAT2025.05 | 2.2 | 1 | 0.02 | |
| GCGModel=Llama3-8B-INST2025.05 | 2.13 | 1.85 | 0.92 | |
| NaiveModel=Llama3-3B-INST2025.05 | 2.05 | 1.46 | 0.54 | |
| WTSModel=Llama3-8B-INST2025.05 | 2.05 | 2.23 | 1.27 | |
| WTSModel=Llama3-3B-INST2025.05 | 1.97 | 1.81 | 0.85 | |
| LINTModel=Llama3-3B-INST2025.05 | 1.77 | 2.69 | 1.12 | |
| BaseModel=Llama3-3B-INST2025.05 | 1.73 | 1.58 | 0.46 | |
| WTSModel=Llama2-7B-CHAT2025.05 | 1.67 | 1.62 | 0.42 | |
| GCGModel=Llama2-7B-CHAT2025.05 | 1.35 | 1.38 | 0.42 | |
| BaseModel=Llama2-7B-CHAT2025.05 | 1.32 | 1.04 | 0.15 |