itercomp-iterative-composition-aware — feedback-learning-from-model-gallery-for

![rw-book-cover](https://d34adp677peecb.cloudfront.net/static/images/article4.6bc1851654a0.png) ## Metadata - Author: [[arxiv.org]] - Full Title:: ITERCOMP: ITERATIVE COMPOSITION-AWARE FEEDBACK LEARNING FROM MODEL GALLERY FOR TEXT-TO-IMAGE GENERATION - Category:: #🗞️Articles - URL:: https://arxiv.org/pdf/2410.07171 - Read date:: [[2026-08-13]] ## Highlights Probably doesn’t apply to Nano Banana and similar models > these models often struggle to follow complex prompts to achieve precise compositional generation (Omost-Team, 2024; Yang et al., 2024b; Zhang et al., 2024b), which requires the model to possess robust, comprehensive capabilities in various aspects, such as attribute binding, spatial relationships, and non-spatial relationships ([View Highlight](https://read.readwise.io/read/01kzx6t491hpr8ea5bcyf66has)) > ![](https://readwise-assets.s3.amazonaws.com/media/reader/pub/050e4aa371a7536e60c8fc379f93dbf4_9QhkhyJ.png?t=1786613018383) ([View Highlight](https://read.readwise.io/read/01kzx6y75kvz3n7nt69j8s4yyd))