Jailbreak Attack on DangerousQA
1.01Harmful RateInitial
Evaluation results
| Method | Links | ||
|---|---|---|---|
| InitialVictim Model=Llama-2-7b-chat2025.02 | 1.01 | 4.5 | |
| RedEvalVictim Model=Llama-2-7b-chat2025.02 | 1.01 | 0.5 | |
| InitialVictim Model=GPT-3.5-turbo-01252025.02 | 1.04 | 7.5 | |
| RedEvalVictim Model=GPT-3.5-turbo-01252025.02 | 3.21 | 45.5 | |
| R2JVictim Model=Llama-2-7b-chat, Avg. Training FLOPS=17396 GFLOPS2025.02 | 4.06 | 90.5 | |
| R2JVictim Model=GPT-3.5-turbo-0125, Avg. Training FLOPS=17396 GFLOPS2025.02 | 4.1 | 91.5 |