PEEK: One-Step-Look-Ahead Exposure Bias Correction for Diffusion Sampling

查看原文 / View Original 编辑 / Edit

Computer Science > Computer Vision and Pattern Recognition arXiv:2611.05407 (cs) [Submitted on 20 Oct 2026] Title: PEEK: One-Step-Look-Ahead Exposure Bias Correction for Diffusion Sampling Authors: Yunwei Bai, Ying Kiat Tan, Tsuhan Chen.

计算机科学 > 计算机视觉与模式识别 arXiv:2611.05407 (cs) [提交于 2026 年 10 月 20 日] 标题：PEEK：用于扩散采样的单步前瞻性曝光偏差校正作者：Yunwei Bai, Ying Kiat Tan, Tsuhan Chen。

Abstract: Diffusion image generators are successful due to accurate denoising networks and strong numerical samplers. Yet the two components are not fully matched. During sampling, the denoiser predicts for the next step based on flawed input that differ much from the real data. This train-test gap, or exposure bias, caps image generation quality (i.e., blurry, lacking in details or awkward in shape).

摘要：扩散图像生成器的成功归功于精确的去噪网络和强大的数值采样器。然而，这两个组件并未完全匹配。在采样过程中，去噪器基于与真实数据差异巨大的缺陷输入来预测下一步。这种训练与测试之间的差距，即曝光偏差，限制了图像生成的质量（例如，图像模糊、细节缺失或形状怪异）。

To mitigate the problem, we propose PEEK, a novel training-free deterministic sampler plugin. At each eligible denoising step, PEEK first performs a provisional one-step update, evaluates the denoiser once at the look-ahead state, and interpolates it with the one-step-later prediction to form the next noisy state. By taking the additional step, we obtain a generally more-image-like prediction from the higher-signal noisy state, mitigating the exposure bias.

为了缓解这一问题，我们提出了 PEEK，这是一种无需训练的确定性采样器插件。在每个符合条件的去噪步骤中，PEEK 首先执行临时的单步更新，在前瞻状态下对去噪器进行一次评估，并将其与后续的单步预测进行插值，以形成下一个噪声状态。通过采取这一额外步骤，我们从高信号噪声状态中获得了更接近图像的预测结果，从而减轻了曝光偏差。

To optimize the interpolation positions, PEEK searches for an interpolation schedule search made efficient by the prefix and baseline noise-schedule-rebasing caches. On 8 pixel-space and latent-space diffusion models with unconditional, label-guided, and text-guided settings across various output resolutions, PEEK consistently improves over NFE-matched baselines, with FID improvements up to 49.3% at minimal extra computation cost (e.g., the one-time-per-model calibration takes at most 2.2 minutes across our experiments.).

为了优化插值位置，PEEK 搜索了一种插值方案，该搜索通过前缀和基准噪声调度重置缓存变得高效。在 8 个像素空间和潜在空间扩散模型上，针对无条件、标签引导和文本引导设置以及各种输出分辨率，PEEK 在 NFE 匹配的基准测试中表现出持续的改进，FID 提升高达 49.3%，且额外计算成本极低（例如，在我们的实验中，每个模型的单次校准最多只需 2.2 分钟）。