用於一步式視覺生成之表徵分佈匹配

摘要

我們闡明了表徵分佈匹配（RDM）的設計空間，此為我們對一種典範的命名，該典範透過在凍結的預訓練編碼器下，比對生成與參考的特徵分佈，來訓練單步影像生成器。我們確立了兩個設計軸向：分佈如何進行比較，以及它們在何種表徵下進行比較。針對這些軸向進行的控制實驗得出了三項發現。首先，經典的最大均值差異（MMD）在十年前無法訓練出令人信服的生成器，一旦正確估計，便成為一個強大且可擴展的目標函數。其次，生成的批次成為關鍵變數，其最優值高於2048，遠超常規的批次大小。第三，任何單一表徵都可能被操縱——生成影像在視覺上仍明顯為假，但其分數卻已低於真實分數——因此我們針對一個平衡的多編碼器組合進行比對，並使用SW_r14（一種基於14個編碼器的切片沃瑟斯坦距離）進行評估，該指標獨立於訓練損失且能抵抗操縱。結合最優選擇後，我們得到改進型RDM（iRDM）：它在ImageNet上以SW_r14 1.30達成了單步生成的最佳表現，並由PickScore（一種從未經我們目標函數優化的人類偏好代理指標）佐證——在71.2%的匹配樣本中，該指標偏好iRDM優於先前最佳的單步生成器。相同的配方將四步驟的FLUX.2 [klein] 後訓練為單步生成器，在GenEval（0.826 對 0.794）和PickScore（22.76 對 22.58）上均超越了四步驟版本，僅耗費90個H200 GPU小時。專案頁面：https://alan-lanfeng.github.io/rdm/。

English

We elucidate the design space of Representation Distribution Matching (RDM), our name for the paradigm that trains a one-step image generator by matching generated and reference feature distributions under frozen pretrained encoders. We identify two design axes, how the distributions are compared and the representations they are compared in, and controlled studies along them yield three findings. First, the classical MMD, which could not train convincing generators a decade ago, becomes a strong and scalable objective once estimated right. Second, the generated batch is then the operative variable, with an optimum above 2048, far beyond customary batch sizes. Third, any single representation can be gamed, driven below the real score while images stay visibly fake, so we match against a balanced battery of encoders and evaluate with SW_r14, a Sliced-Wasserstein distance over 14 encoders that is independent of the training loss and resists gaming. Combining the preferred choices yields improved RDM (iRDM): it sets the one-step state of the art on ImageNet at SW_r14 1.30, corroborated by PickScore, a human-preference proxy our objective never optimizes, which prefers it over the prior best one-step generator on 71.2% of matched samples. The same recipe post-trains the four-step FLUX.2 [klein] into a one-step generator, surpassing the four-step version on GenEval, 0.826 to 0.794, and on PickScore, 22.76 to 22.58, in 90 H200 GPU-hours. Project page: https://alan-lanfeng.github.io/rdm/.