面向图像生成的多尺度局部推测解码
Multi-Scale Local Speculative Decoding for Image Generation
把LLM中的推测解码迁移到自回归图像生成，用低分辨率drafter提出候选token、高分辨率target并行验证，并引入局部拒绝重采样替代光栅扫描式重采样，可直接用于加速自回归视觉生成模型的采样。
原文依据 a low-resolution drafter paired with an up-sampling step to propose candidate image tokens, which are then verified in parallel by a high-resolution target model
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论自回归（AR）模型在图像合成中取得了显著成功，但其序列化特性带来了严重的延迟约束。推测解码为加速提供了一条有前景的途径，但现有方法受限于token级歧义且缺乏空间感知。本文提出多尺度局部推测解码（MuLo-SD），一种将多分辨率草稿与空间感知验证相结合的新框架，用于加速自回归图像生成。该方法利用低分辨率草稿器配合上采样步骤提出候选图像token，随后由高分辨率目标模型并行验证。关键在于，我们引入了局部拒绝与重采样机制，通过在首次拒绝后聚焦空间邻域而非光栅扫描式重采样，实现对草稿错误的高效纠正。当与并行解码重采样结合时，MuLo-SD 实现了显著加速——最高达5倍——在加速比上优于推测解码与并行解码基线，同时保持可比的语义对齐与感知质量。这些结果在 MS-COCO 5k 验证集上通过 GenEval、DPG-Bench 以及 FID/HPSv2 得到验证。大量消融实验凸显了上采样设计、概率池化以及带邻域扩展的局部拒绝与重采样的影响。该方法在图像合成的推测解码方面树立了新的最优水平，弥合了效率与保真度之间的差距。
英文原文摘要 1578 字符
Autoregressive (AR) models have achieved remarkable success in image synthesis, yet their sequential nature imposes significant latency constraints. Speculative Decoding offers a promising avenue for acceleration, but existing approaches are limited by token-level ambiguity and lack of spatial awareness. In this work, we introduce Multi-Scale Local Speculative Decoding (MuLo-SD), a novel framework that combines multi-resolution drafting with spatially informed verification to accelerate AR image generation. Our method leverages a low-resolution drafter paired with an up-sampling step to propose candidate image tokens, which are then verified in parallel by a high-resolution target model. Crucially, we incorporate a local rejection and resampling mechanism, enabling efficient correction of draft errors by focusing on spatial neighborhoods rather than raster-scan resampling after the first rejection. When integrated with parallel decoding resampling, \methodname achieves substantial speedups -- up to 5x -- outperforming both speculative decoding and parallel decoding baselines in terms of acceleration, while maintaining comparable semantic alignment and perceptual quality. These results are validated using GenEval, DPG-Bench, and FID/HPSv2 on the MS-COCO 5k validation split. Extensive ablations highlight the impact of up-sampling design, probability pooling, and local rejection and resampling with neighborhood expansion. Our approach sets a new state-of-the-art in speculative decoding for image synthesis, bridging the gap between efficiency and fidelity.