Rethinking Prompt Design for Inference-time Scaling in Text-to-Visual Generation

Subin Kim, Sangwoo Mo, Mamshad Nayeem Rizve +4

Achieving precise alignment between user intent and generated visuals remains a central challenge in text-to-visual generation, as a single attempt often fails to produce the desir…