无需假评分网络的分布匹配蒸馏
Distribution Matching Distillation without Fake Score Network
AI总结：
本文提出无需假评分网络的分布匹配蒸馏（FSF-DMD），通过流图生成器自身诱导的伪速度替代传统假评分网络，实现了分布级校正，并在ImageNet-1K数据集上验证了其有效性。
摘要：
分布匹配蒸馏（DMD）为少步生成提供了有效的分布级校正，但依赖辅助的假评分网络来跟踪生成分布的演变。近期工作将DMD式目标与流图生成器结合，以利用正向发散训练和反向发散校正。假评分估计器仍是一个额外的组件，具有内存和更新开销。在本工作中，我们研究当生成器本身具有流图结构时是否可以避免显式跟踪器。我们提出无需假评分网络的DMD（FSF-DMD），一种适用于流图生成器的DMD形式，其用生成器诱导的伪速度替代传统假评分估计器。关键观察是流图生成器的端点伪速度提供了一个可计算的假速度估计代理，使生成器本身能够提供反向发散信号。基于这一观察，我们推导出一个实用的目标，扩展了流图一致的反向模拟，并引入了自教师变体以从头开始训练。在ImageNet-1K 256×256实验中，FSF-DMD改进了流图基线，达到了流图初始化设置下低于列出的DMD2比较的FID，并在流图匹配初始化和从头开始训练时仍保持有效。
英文摘要：
Distribution Matching Distillation (DMD) provides an effective distribution-level correction for few-step generation, while relying on an auxiliary fake-score network to track the evolving generative distribution. Recent work combines DMD-style objectives with flow-map generators to exploit both forward-divergence training and reverse-divergence correction. The fake-score estimator remains an additional component with memory and update overhead. In this work, we study whether this explicit tracker can be avoided when the generator itself has a flow-map structure. We propose Fake-Score-network-Free DMD (FSF-DMD), a DMD formulation for flow-map generators that replaces the auxiliary fake-score estimator with a generator-induced pseudo-velocity surrogate. The key observation is that the endpoint pseudo-velocity of a flow-map generator provides a tractable proxy for fake-velocity estimation, allowing the generator itself to supply the reverse-divergence signal. Building on this observation, we derive a practical objective, extend it with flow-map-consistent backward simulation, and introduce a self-teacher variant for training from scratch. In our ImageNet-1K $256 \times 256$ experiments, FSF-DMD improves flow-map baselines, reaches lower FID than the listed DMD2 comparisons in the flow-map-initialized setting, and remains effective under flow-matching initialization and training from scratch.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
