Author name: Xinyi Wen, Lena Hegemann, Xiaofu Jin, Shuai Ma, Antti Oulasvirta

arXiv:2602.04713v2 Announce Type: replace-cross
Abstract: Aligning text-to-image generation with user intent remains challenging, as users frequently provide ambiguous inputs and struggle with model idiosyncrasies. We propose Adaptive Prompt Elicitati…