Experimental Data

[kaikaikai7]

attack-artifacts/PAIR/black_box/gpt-4-0125-preview.json Is the victim model of this experimental data gpt-4-0125-preview or vicuna

[kaikaikai7]

Do "total_prompt_tokens" and "total_response_tokens" only include the victim model, or do they include the attack model and the evaluation model?And what is the difference between "number_of_queries" and "queries_to_jailbreak"?