用分层奖励优化文本生成，让模型更准地渲染文字。
TextAlign: Preference Alignment for Text Rendering with Hierarchical Rewards
- 设计分层视觉语言奖励，从全局到字形逐级评估文本错误。
- 在FLUX.1-dev和Z-Image-Turbo上提升OCR准确率，不损害图像质量。
- 无需修改模型结构，适合各类基础生成模型快速适配。
忠实的文本渲染仍是大文本到图像生成模型的顽疾，因其需同时遵循语义指令并保持精细的字形结构。以往方法多通过特定架构模块或编码器修改来改进，但增加跨基础模型部署的复杂性。本文将文本渲染视为后训练阶段的偏好对齐问题，提出非侵入式框架TextAlign，保持生成器架构不变。核心是基于分层视觉语言模型（VLM）的奖励机制，将渲染误差分解为全局、单词和字形三个层级，并将二值缺陷判断转化为标量偏好信号。该信号支持组相对策略优化（GRPO）与直接偏好优化（DPO）。在FLUX.1-dev与Z-Image-Turbo上的实验显示，文本渲染的OCR准确率持续提升，且未降低整体生成质量。相较于包括SD3.5、Qwen-Image、AnyText和TextDiffuser在内的强基线模型，结果表明奖励设计是改进文本渲染的可扩展替代方案，优于模型重设计。
原文摘要 · Abstract (English)
Faithful text rendering remains a persistent weakness of large text-to-image generative models, as it requires both semantic instruction following and fine-grained glyph-level structure. Prior methods often improve this ability through architecture-specific modules or encoder modifications, which complicate deployment across foundation models. We study text rendering as a post-training preference-alignment problem and propose TextAlign, a non-invasive framework that keeps the generator architecture unchanged. The key component is a hierarchical vision-language model (VLM)-based reward that decomposes rendering errors into global, word, and glyph levels, then converts binary defect judgments into a scalar preference signal. The resulting signal supports both Group Relative Policy Optimization (GRPO) and Direct Preference Optimization (DPO). Experiments on FLUX.1-dev and Z-Image-Turbo show consistent gains in OCR-based text accuracy without degrading general generation quality. Compared with strong foundation and text-rendering baselines, including SD3.5, Qwen-Image, AnyText, and TextDiffuser, these results indicate that reward design offers a scalable alternative to model redesign for improving text rendering.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。