发表机构
AITHYRA; University of Oxford; Mila - Québec AI Institute; Université de Montréal; Massachusetts Institute of Technology; Technische Universität Wien; Flatiron Institute; University of Alberta; Alberta Machine Intelligence Institute（AITHYRA; 牛津大学; Mila - 魁北克人工智能研究所; 蒙特利尔大学; 麻省理工学院; 维也纳工业大学; 熨斗研究所; 阿尔伯塔大学; 阿尔伯塔机器智能研究所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对条件流模型，提出全局类别无关的最优传输耦合GT，与无分类器引导结合时跨领域和规模提升生成质量，为训练时性能优化提供新方向。
AI 中文摘要
最优传输耦合已被证明能够降低无条件流模型中的训练方差，但其在条件生成中的作用仍不清楚。一种自然的方法是为每个条件分别构建耦合，但这对于现代图像基础模型中常见的大规模或连续条件空间而言并不实际。我们引入了全局传输（GT），一种全局的、与类别无关的最优传输耦合，其计算无需类别标签。GT 可以将不同的条件与源噪声的不同区域相关联，因此在无引导情况下会降低性能。然而，当与无分类器引导（CFG）结合时，GT 在多个领域、模型规模和采样预算下持续提升生成质量。这一反转表明，条件流的耦合应在推理时使用的引导流下进行实证和理论评估，而非在无引导生成下评估。我们在离散类别和连续文本条件图像生成中，跨模型规模评估了 GT，并研究了耦合选择如何改变引导轨迹。这些结果将引导流设置中的耦合设计识别为一个简单的训练时间轴，无需修改现有架构、采样器或引导机制即可提升性能。
英文摘要
Optimal-transport couplings have been shown to reduce training variance in unconditional flow models, but their role in conditional generation remains unclear. A natural approach constructs separate couplings for each condition, but this is impractical for large or continuous conditioning spaces found in modern image foundation models. We introduce Global Transport (GT), a global class-agnostic optimal-transport coupling, computed without class labels. GT can associate different conditions with different regions of the source noise, and consequently worsens performance without guidance. However, when combined with classifier-free guidance (CFG), GT consistently improves generation across domains, model scales, and sampling budgets. This reversal suggests that couplings for conditional flows should be evaluated both empirically and theoretically under the guided flow used at inference, rather than on unguided generation. We evaluate GT over both discrete class and continuous text conditioned image generation across model scales, and investigate how coupling choice alters guided trajectories. These results identify coupling design in the guided flow setting as a simple training time axis to improve performance without modifying existing architectures, samplers, or guidance mechanisms.
