Targeted Attack on GPT closed-source standard MLLMs 5.2
3.3ASRFRA-Attack
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FRA-AttackThreshold=0.92026.05 | 3.3 | 0.427 | |
| M-Attack-V2Threshold=0.92026.05 | 2.2 | 0.383 | |
| FOA-AttackThreshold=0.92026.05 | 0.8 | 0.312 | |
| M-AttackThreshold=0.92026.05 | 0.6 | 0.272 | |
| AttackVLMThreshold=0.92026.05 | 0 | 0.06 | |
| AdvDiffVLMThreshold=0.92026.05 | 0 | 0.047 | |
| SSA-CWAThreshold=0.92026.05 | 0 | 0.049 | |
| AnyAttackThreshold=0.92026.05 | 0 | 0.053 |