This highlights the current imbalance in visual understanding and generation capabilities of unified multimodal models: they ...