面向文本到图像生成的代理修复器
Agentic Retoucher for Text-To-Image Generation
AI总结：
本文提出Agentic Retoucher框架，通过感知-推理-行动循环改进文本到图像生成的质量，引入GenBlemish-27K数据集进行评估，提升图像真实感与局部修正可靠性。
摘要：
文本到图像（T2I）扩散模型如SDXL和FLUX已实现卓越的逼真度，但四肢、面部、文本等部位仍存在微小畸变。现有修复方法或成本高昂的迭代重生成，或依赖空间定位弱的视觉语言模型，导致语义漂移和不可靠的局部编辑。为弥合这一差距，我们提出Agentic Retoucher，一种分层决策驱动框架，将生成后修正重新表述为人类感知-推理-行动循环。具体而言，我们设计（1）一个感知代理，学习上下文显著性以在文本图像一致性提示下实现细粒度畸变定位；（2）一个推理代理，通过渐进性偏好对齐进行人类对齐的推断诊断；（3）一个行动代理，根据用户偏好自适应规划局部修复。此设计将感知证据、语言推理和可控修正整合到统一的自我修正决策过程中。为实现细粒度监督和定量评估，我们进一步构建GenBlemish-27K数据集，包含6K T2I图像和27K标注的缺陷区域，覆盖12类。广泛实验表明，Agentic Retoucher在感知质量、畸变定位和人类偏好对齐方面均优于现有最先进方法，确立了自我修正和感知可靠的T2I生成新范式。
英文摘要：
Text-to-image (T2I) diffusion models such as SDXL and FLUX have achieved impressive photorealism, yet small-scale distortions remain pervasive in limbs, face, text and so on. Existing refinement approaches either perform costly iterative re-generation or rely on vision-language models (VLMs) with weak spatial grounding, leading to semantic drift and unreliable local edits. To close this gap, we propose Agentic Retoucher, a hierarchical decision-driven framework that reformulates post-generation correction as a human-like perception-reasoning-action loop. Specifically, we design (1) a perception agent that learns contextual saliency for fine-grained distortion localization under text-image consistency cues, (2) a reasoning agent that performs human-aligned inferential diagnosis via progressive preference alignment, and (3) an action agent that adaptively plans localized inpainting guided by user preference. This design integrates perceptual evidence, linguistic reasoning, and controllable correction into a unified, self-corrective decision process. To enable fine-grained supervision and quantitative evaluation, we further construct GenBlemish-27K, a dataset of 6K T2I images with 27K annotated artifact regions across 12 categories. Extensive experiments demonstrate that Agentic Retoucher consistently outperforms state-of-the-art methods in perceptual quality, distortion localization and human preference alignment, establishing a new paradigm for self-corrective and perceptually reliable T2I generation.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
