Safety Robustness on Random Insertion Attack
2.7ASRInsertion
Evaluation Results
| Method | Links | |
|---|---|---|
| InsertionBackbone=Llama-2-7b-Chat, Harmful span length=252026.06 | 2.7 | |
| InsertionBackbone=Llama-2-7b-Chat, Harmful span length=502026.06 | 3.4 | |
| InsertionBackbone=Llama-2-7b-Chat, Harmful span length=752026.06 | 4.8 | |
| InsertionBackbone=Llama-2-7b-Chat, Harmful span length=1002026.06 | 5.9 | |
| PrefillBackbone=Llama-2-7b-Chat, Harmful span length=252026.06 | 18.4 | |
| PrefillBackbone=Llama-2-7b-Chat, Harmful span length=502026.06 | 22.3 | |
| PrefillBackbone=Llama-2-7b-Chat, Harmful span length=752026.06 | 27.6 | |
| PrefillBackbone=Llama-2-7b-Chat, Harmful span length=1002026.06 | 34.3 | |
| CleanBackbone=Llama-2-7b-Chat, Harmful span length=252026.06 | 76.6 | |
| CleanBackbone=Llama-2-7b-Chat, Harmful span length=502026.06 | 80.4 | |
| CleanBackbone=Llama-2-7b-Chat, Harmful span length=752026.06 | 83.6 | |
| CleanBackbone=Llama-2-7b-Chat, Harmful span length=1002026.06 | 85.8 |