HuggingFace Daily Papers（社区热门论文）· HuggingFace Daily Papers（社区热门论文）·· 2026-07-02精选AI 评分74
表示分布匹配（RDM）用于一步视觉生成
AI 导读
表示分布匹配（RDM）通过匹配冻结预训练编码器下的生成与参考特征分布来训练一步图像生成器。三个发现：经典MMD正确估计后成为可扩展目标；生成批次大小最优超过2048；单一表示可被欺骗，需匹配平衡编码器组合并采用独立于训练损失的SW_r14评估。改进的iRDM在ImageNet上以SW_r14 1.30达一步生成SOTA，PickScore在71.2%样本上偏好iRDM。该方法将四步FLUX.2后训练为一步生成器，在GenEval（0.826对0.794）和PickScore（22.76对22.58）上超越原版，仅需90 H200 GPU小时。
推荐理由
这篇论文把MMD重新变成了一流的一步生成目标，用多编码器匹配防止作弊，并把FLUX.2四步模型浓缩成一步，性能不降反升，90 GPU小时就能复现，做视觉生成的人应该仔细看。
正文
Abstract:We elucidate the design space of Representation Distribution Matching (RDM), our name for the paradigm that trains a one-step image generator by matching generated and reference feature distributions under frozen pretrained encoders. We identify two design axes, how the distributions are compared and the representations they are compared in, and controlled studies along them yield three findings. First, the classical MMD, which could not train convincing generators a decade ago, becomes a strong and scalable objective once estimated right. Second, the generated batch is then the operative variable, with an optimum above 2048, far beyond customary batch sizes. Third, any single representation can be gamed, driven below the real score while images stay visibly fake, so we match against a balanced battery of encoders and evaluate with SW_r14, a Sliced-Wasserstein distance over 14 encoders that is independent of the training loss and resists gaming. Combining the preferred choices yields improved RDM (iRDM): it sets the one-step state of the art on ImageNet at SW_r14 1.30, corroborated by PickScore, a human-preference proxy our objective never optimizes, which prefers it over the prior best one-step generator on 71.2% of matched samples. The same recipe post-trains the four-step FLUX.2 [klein] into a one-step generator, surpassing the four-step version on GenEval, 0.826 to 0.794, and on PickScore, 22.76 to 22.58, in 90 H200 GPU-hours. Project page: this https URL.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2607.02375 [cs.CV] |
| (or arXiv:2607.02375v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.02375 arXiv-issued DOI via DataCite |
Submission history
From: Lan Feng [view email]
[v1]
Thu, 2 Jul 2026 16:15:38 UTC (14,821 KB)
来源：HuggingFace Daily Papers（社区热门论文） · arxiv.org