1 paper
Wanguang Li, Zhaoxin Wang, Handing Wang
Text-to-image (T2I) systems typically have prompt-level safety filters before the generator to block unsafe requests, yet such systems remain vulnerable to malicious jailbreak prom…