search

cure-lab / MMA-Diffusion

[CVPR2024] MMA-Diffusion: MultiModal Attack on Diffusion Models
Other
151 stars 7 forks source link

Some questions regarding text-based modality attacks. #15

Open yszbb opened 1 month ago

yszbb commented 1 month ago

Thank you very much for your excellent work. However, I have a question regarding text-based modality attacks. It appears that open-source T2I models do not have prompt filters deployed. Therefore, does the data in Table 1 only assess whether adversarial prompts can generate images with NSFW content?

yangyijune commented 1 month ago

Yes, your understanding is correct.