ResearchBenchmarksAdversarial Attack on Text-to-Image Generation on ImageNet Long Prompt 1K (val)For AI assistantsMarkdownContent indexAdversarial Attack on Text-to-Image Generation on ImageNet Long Prompt 1K (val)Follow17.95FIDClean17.48220.64123.826.959Jun 5, 2023Evaluation resultsMethodMethodLinksFIDAttack Success RateISTSCleanPrompt Type=LongPrompt Type=Long2023.0617.95——1RandomPrompt Type=LongPrompt Type=Long2023.0624.1641.4—0.94ATMPrompt Type=Long, Sear...Prompt Type=Long, Search iterations=100, Learning rate=0.3, Margin=302023.0629.6581.2—0.84