Figureure 8 · Qualitative comparison of image editing resultsFigure 8. Qualitative comparison of image editing results. While existing methods frequently modify incorrect objects or apply changes broadly across the scene, our method successfully localizes the edit to the intended target. This demonstrates that providing explicit spatial conditioning enables more precise and reliable edit localization. Additional qualitative comparisons are provided in Figure 11.这张图/表用于判断 Analysis-by-Proxy 的实验收益来自哪里。重点看替换、消融或跨模型设置下趋势是否一致，而不是只看单个最高分。“Recolour the bottom right clownfish to be black and white” “Make the bottom jellyfish glo“Recolour the bottom right clownfish to be black and white” “Make the bottom jellyfish glow red” Figure 10. Green: the explicit bounding box prediction generated by the trained Q-Former proxy, which isolates the target region to guide the downstream localized edit. The bottom row shows the corresponding failures of the original Qwen-Image-Edit pipeline on the same inputs.这张图概括 Analysis-by-Proxy 的整体方法流程。阅读时先看模块之间传递的训练信号，再看作者如何把目标拆成可优化的子问题。
