VisionDirector：视觉语言引导的闭环优化用于生成式图像合成
VisionDirector: Vision-Language Guided Closed-Loop Refinement for Generative Image Synthesis
提出一种无需训练的视觉语言监督器，通过结构化目标提取、动态决策、微网格采样与语义验证回滚，实现长指令下的闭环图像生成与编辑；并利用Group Relative Policy Optimization微调规划器，为可控生成提供可复用的规划与验证框架。
原文依据 VisionDirector, a training-free, vision-language supervisor that (i) extracts structured goals from long instructions, (ii) dynamically decides between one-shot generation and staged edits
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论生成模型如今能够生成逼真的图像，但仍难以处理专业设计师提出的长而多目标的提示。为揭示这一差距并更好地评估模型在真实场景中的表现，我们引入了Long Goal Bench（LGBench），一个包含2000个任务的测试套件（1000个文本到图像任务，1000个图像到图像任务），其平均指令包含18至22个紧密耦合的目标，涵盖全局布局、局部物体放置、排版和标志保真度。我们发现，即使是最优的商业API也仅能满足不到72%的目标，并且经常遗漏局部编辑，这证实了当前流程的脆弱性。为解决这一问题，我们提出了VisionDirector，一个无需训练的视觉语言监督器，它（i）从长指令中提取结构化目标，（ii）动态决定采用单样本生成还是分阶段编辑，（iii）在每次编辑后运行微网格采样以及语义验证/回滚，（iv）记录目标级别的奖励。我们进一步使用Group Relative Policy Optimization微调规划器，从而获得更短的编辑轨迹（3.1步对4.2步）和更强的一致性。VisionDirector在GenEval上取得了新的最优结果（总体提升7%），在ImgEdit上也有提升（绝对提升0.07），同时在排版、多物体场景和姿态编辑方面产生了持续的定性改进。代码、基准和评估脚本将发布。
英文原文摘要 1371 字符
Generative models can now produce photorealistic imagery, yet they still struggle with the long, multi-goal prompts that professional designers issue. To expose this gap and better evaluate models' performance in real-world, we introduce Long Goal Bench(LGBench), a 2000-task suite (1000 T2I, 1000 I2I) whose average instruction contains 18---22 tightly coupled goals spanning global layout, local object placement, typography, and logo fidelity. We find even state-of-the-art commercial APIs satisfy fewer than 72% of the goals and routinely miss localized edits, confirming the brittleness of current pipelines. To address this, we present VisionDirector, a training-free, vision-language supervisor that (i) extracts structured goals from long instructions, (ii) dynamically decides between one-shot generation and staged edits, (iii) runs micro-grid sampling plus semantic verification/rollback after every edit, and (iv) logs goal-level rewards. We further fine-tune the planner with Group Relative Policy Optimization, yielding shorter edit trajectories (3.1 vs. 4.2 steps) and stronger alignment. VisionDirector achieves new state of the art on GenEval (+7% overall), and ImgEdit (+0.07 absolute) while producing consistent qualitative improvements on typography, multi-object scenes, and pose editing. The code, benchmark, and evaluation scripts will be released.