发表机构
University of Toronto（多伦多大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出生成感知重建（GAR）方法，通过构建从重建到生成的潜在轨迹来诊断分布不匹配，并利用中间潜在变量的成对监督改进解码器，从而提升生成质量。
AI 中文摘要
在潜在生成模型中，重建质量通常被认为与生成性能相关。然而，重建FID（rFID）可能与生成FID（gFID）呈现弱相关甚至负相关。我们将这一差异归因于潜在分布不匹配：重建评估的是解码器在编码器诱导的潜在变量上的表现，而生成则使用同一解码器处理由生成模型产生的潜在变量。为了刻画这种偏移，我们引入了生成感知重建（GAR），该方法通过向编码器潜在变量添加噪声并通过生成模型去噪后再解码，构建了一条从标准重建到生成的连续轨迹。GAR沿此轨迹探测解码器行为，使从编码器到生成时潜在分布的转变变得可观察和可诊断。基于轨迹的诊断指标GAR-FID在多种分词器（tokenizer）和规模下与gFID表现出强经验相关性。重要的是，中间GAR潜在变量在保持与其源图像对应关系的同时变得更加生成感知，从而保留了完全生成潜在变量所缺失的成对监督。这种对应关系使得能够在中间GAR潜在变量上进行解码器适配，从而在不同模型规模下持续提升生成质量。总体而言，潜在分布不匹配为评估和改进潜在生成模型提供了一个有用的视角。
英文摘要
In latent generative models, reconstruction quality is often assumed to correlate with generative performance. However, reconstruction FID (rFID) can exhibit weak or even negative correlation with generation FID (gFID). We attribute this misalignment to a latent distribution mismatch: reconstruction evaluates the decoder on encoder-induced latents, whereas generation uses the same decoder on latents produced by the generative model. To characterize this shift, we introduce generation-aware reconstruction (GAR), which constructs a continuous trajectory from standard reconstruction toward generation by perturbing encoder latents with noise and denoising them through the generative model before decoding. GAR probes the decoder behavior along this trajectory, making the transition from encoder to generation-time latent distributions observable and diagnosable. The resulting trajectory-based diagnostic, GAR-FID, exhibits strong empirical correlation with gFID across diverse tokenizers and scales. Importantly, intermediate GAR latents become more generation-aware while preserving correspondence with their source images, thereby retaining paired supervision that is absent for fully generated latents. This correspondence enables decoder adaptation on intermediate GAR latents, consistently improving generative quality across model scales. Overall, latent distribution mismatch provides a useful perspective for evaluating and improving latent generative models.
Comments27 pages, 23 figures,and 15 tables
