JLT：潛在擴散變換器中的乾淨潛在預測

摘要

以乾淨資料預測進行流匹配已展現出，相較於預測環境噪聲量，回歸乾淨點更能有效利用低維度結構。我們探討此原則在影像映射至學習後的潛在空間（其中壓縮已去除大量原始像素變異性）後是否仍具效用。我們引入JLT，這是一個基於凍結FLUX.2 VAE編碼的1.3億參數潛在擴散Transformer，並在同一表徵、骨幹網路及訓練設定下，比較乾淨潛在預測與匹配的速度預測DiT。儘管三個變數x、ε及v在固定破壞時間下可線性轉換，但局部高斯分析顯示，速度回歸會繼承各向同性目標共變異數下限，並放大低變異潛在方向，而乾淨預測則抑制之。在ImageNet 256×256上，JLT-B/1使用無分類器引導獲得FID-50K 2.50，並與速度預測存在顯著的匹配目標差距。這些結果表明，潛在擴散中的預測目標是依表徵而定的幾何選擇，而非可互換的代數參數化。

English

Flow matching with clean-data prediction has shown that regressing the clean point can exploit low-dimensional structure more effectively than predicting an ambient noised quantity. We ask whether this principle remains useful after images are mapped into a learned latent space, where compression has already removed much of the raw pixel variability. We introduce JLT, a 130M latent diffusion Transformer over frozen FLUX.2 VAE codes, and compare clean-latent prediction with a matched velocity-prediction DiT under the same representation, backbone, and training settings. Although the three variables x, epsilon, and v are linearly convertible for a fixed corruption time, a local Gaussian analysis shows that velocity regression inherits an isotropic target-covariance floor and amplifies low-variance latent directions, while clean prediction damps them. On ImageNet 256 x 256, JLT-B/1 obtains FID-50K 2.50 with classifier-free guidance, with a large matched-target gap over velocity prediction. These results suggest that prediction targets in latent diffusion are representation-dependent geometric choices, rather than interchangeable algebraic parameterizations.