1 paper · 1 filter
Mingxin Liu, Ziqian Fan, Zhaokai Wang +13
Unified multimodal models target joint understanding, reasoning, and generation, but current image editing benchmarks are largely confined to natural images and shallow commonsense…