1 paper · 1 filter
Xindi Wu, Dingli Yu, Yangsibo Huang +2
Compositionality is a critical capability in Text-to-Image (T2I) models, as it reflects their ability to understand and combine multiple concepts from text descriptions. Existing e…