{"slug": "peek-one-step-look-ahead-exposure-bias-correction-for-diffusion-sampling", "title": "PEEK: One-Step-Look-Ahead Exposure Bias Correction for Diffusion Sampling", "summary": "Researchers introduced PEEK, a training-free deterministic sampler plugin that corrects exposure bias in diffusion image generators by performing a provisional one-step update, evaluating the denoiser at the look-ahead state, and interpolating it with the one-step-later prediction to form the next noisy state. Across 8 pixel-space and latent-space diffusion models in unconditional, label-guided and text-guided settings, PEEK improved FID by up to 49.3% over NFE-matched baselines at minimal extra computation cost, with one-time-per-model calibration taking at most 2.2 minutes. The work, posted as arXiv:2611.05407v1, targets the train-test gap that caps image quality, producing blurry, detail-poor or awkwardly shaped outputs.", "body_md": "arXiv:2611.05407v1 Announce Type: new \nAbstract: 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). 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. 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.). Project website is at https://dachengruoque.github.io/PEEK-Video/.", "url": "https://wpnews.pro/news/peek-one-step-look-ahead-exposure-bias-correction-for-diffusion-sampling", "canonical_source": "https://arxiv.org/abs/2611.05407", "published_at": "2026-11-05 05:00:00+00:00", "updated_at": "2026-11-05 05:23:06.778014+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "generative-ai", "computer-vision", "ai-research"], "entities": ["PEEK", "arXiv", "FID"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/peek-one-step-look-ahead-exposure-bias-correction-for-diffusion-sampling", "markdown": "https://wpnews.pro/news/peek-one-step-look-ahead-exposure-bias-correction-for-diffusion-sampling.md", "text": "https://wpnews.pro/news/peek-one-step-look-ahead-exposure-bias-correction-for-diffusion-sampling.txt", "jsonld": "https://wpnews.pro/news/peek-one-step-look-ahead-exposure-bias-correction-for-diffusion-sampling.jsonld"}}