Question Answering on MS MARCO (150 random samples)
100BAROriginal LLM
Evaluation Results
| Method | Links | |
|---|---|---|
| Original LLMAttack=Adv Strings, Model Backbone=Vicuna-7B-v1.32023.09 | 100 | |
| Original LLMAttack Type=AutoDAN, Base Model=Vicuna-7B-chat-HF2023.09 | 99.3 | |
| Original LLMAttack Type=AutoDAN-GPT, Base Model=Vicuna-7B-chat-HF2023.09 | 99.3 | |
| Original LLMAttack Type=TAP, Base Model=Vicuna-7B-chat-HF2023.09 | 99.3 | |
| RA-LLMAttack=Adv Strings, Model Backbone=Vicuna-7B-v1.32023.09 | 99 | |
| RA-LLMAttack Type=AutoDAN, Base Model=Vicuna-7B-chat-HF2023.09 | 98.7 | |
| RA-LLMAttack Type=AutoDAN-GPT, Base Model=Vicuna-7B-chat-HF2023.09 | 98.7 | |
| RA-LLMAttack Type=TAP, Base Model=Vicuna-7B-chat-HF2023.09 | 98.7 | |
| Original LLMAttack Type=AutoDAN, Base Model=Guanaco-7B-HF2023.09 | 95.3 | |
| Original LLMAttack Type=AutoDAN-GPT, Base Model=Guanaco-7B-HF2023.09 | 95.3 | |
| Original LLMAttack Type=TAP, Base Model=Guanaco-7B-HF2023.09 | 95.3 | |
| RA-LLMAttack Type=AutoDAN, Base Model=Guanaco-7B-HF2023.09 | 92 | |
| RA-LLMAttack Type=AutoDAN-GPT, Base Model=Guanaco-7B-HF2023.09 | 92 | |
| RA-LLMAttack Type=TAP, Base Model=Guanaco-7B-HF2023.09 | 92 |