4 ms·
Glaze: Protecting Artists from Style Mimicry by Text-to-Image Models
- josephcsible 4y agoHow can text-to-image models be patched against the vulnerability that the attack in this paper exploits?
- smoldesu 4y agoTraining with older unpoisoned datasets seems like a good start.