Beyond the Estimate: Wiener-Informed Decoding for Easier Training of Pixel-Space Flow Models
Computer Science > Computer Vision and Pattern Recognition arXiv:2610.36192 (cs) [Submitted on 26 Oct 2026] Title: Beyond the Estimate: Wiener-Informed Decoding for Easier Training of Pixel-Space Flow Models Authors: Ziqiao Wang, Yang You.
计算机科学 > 计算机视觉与模式识别 arXiv:2610.36192 (cs) [提交于 2026 年 10 月 26 日] 标题：超越估计：用于简化像素空间流模型训练的维纳信息解码（Wiener-Informed Decoding）。作者：Ziqiao Wang, Yang You。
Abstract: Learning velocity fields directly in pixel space is difficult and slow. We show that coarse second-order image statistics can ease this task: frozen Wiener estimates improve velocity learning, with gains extending beyond the linear predictions they explicitly provide. Moreover, we observe similar but complementary gains from frozen bases and pixel decoding.
摘要：直接在像素空间学习速度场既困难又缓慢。我们证明，粗略的二阶图像统计数据可以简化这一任务：冻结的维纳估计（Wiener estimates）改善了速度学习，其收益超出了它们明确提供的线性预测范围。此外，我们观察到冻结基底和像素解码也带来了类似但互补的收益。
These observations motivate Wiener-informed Decoding (WiD), which uses statistical knowledge both for direct prediction and to organize decoder computation without modifying native blocks. WiD combines a trainable Wiener prediction path with coefficient interfaces that inject features into the native decoder and return its predicted coefficient corrections, with these components initialized with data statistics.
这些观察结果促成了维纳信息解码（WiD），它利用统计知识进行直接预测，并组织解码器计算，而无需修改原生模块。WiD 将可训练的维纳预测路径与系数接口相结合，将特征注入原生解码器并返回其预测的系数校正值，这些组件均使用数据统计信息进行初始化。
Across model scales and decoder architectures, WiD accelerates training and improves generation quality with modest parameter overhead. It also yields smoother gradient trajectories and remains effective with representation alignment. WiD improves class-conditional and text-to-image generation across different resolutions.
在各种模型规模和解码器架构中，WiD 以适度的参数开销加速了训练并提高了生成质量。它还产生了更平滑的梯度轨迹，并在表征对齐方面保持有效。WiD 改善了不同分辨率下的类条件生成和文本到图像生成效果。
On ImageNet 256 x 256, PixelDiT-XL with WiD reaches FID 1.53 after 500 epochs, achieving comparable quality to PixelDiT with 37.5% fewer training epochs. Statistical denoisers contribute beyond the estimate: their computational structure can also ease pixel-space training.
在 ImageNet 256 x 256 数据集上，使用 WiD 的 PixelDiT-XL 在 500 个 epoch 后达到了 1.53 的 FID，在训练 epoch 减少 37.5% 的情况下实现了与 PixelDiT 相当的质量。统计去噪器的贡献不仅限于估计本身：它们的计算结构也可以简化像素空间的训练。