[论文] [论文] Expanding Flow Maps

论文概要

研究领域: ML 作者: Sophia Tang, Pranam Chatterjee 发布时间: 2026-07-24 arXiv: 2507.19317

中文摘要

基于流的生成模型在连续和离散状态空间的快速可控生成方面取得了显著进展，但现有参数化受限于固定维度或固定序列长度。在此，我们引入了扩展生成流（EFlows），它定义了沿扩展插值分布之间维数递增的流，通过用条件噪声增强状态来增长状态。基于这一构建，我们提出了扩展流映射（EFMs），一类新的流映射，将扩展插值蒸馏为高效的少步生成模型。每个EFM将任意两个时间步之间的映射分解为两个可学习操作：扩展操作，基于当前状态用新坐标或Token增强状态空间；以及传输映射，将扩展状态沿插值向前推进。组合这些操作得到一个联合扩展和去噪状态的单一映射，恢复现有固定画布流和流映射作为扩展操作为恒等映射的特例。我们进一步将该框架扩展到离散单纯形，实现可变大小图生成和可变长度序列生成。在连续和离散模态中，我们将EFlows和EFMs确立为输出大小本身是可学习的、可控的自由度的有原则框架。

原文摘要

Flow-based generative models have enabled remarkable progress in fast and controllable generation across continuous and discrete state spaces, yet existing parameterizations are constrained to fixed dimensions or fixed sequence lengths. Here, we introduce Expanding Generative Flows (EFlows), which define flows between distributions of increasing dimensionality along an expanding interpolant that grows the state by augmenting it with conditional noise. Building on this construction, we propose Expanding Flow Maps (EFMs), a new class of flow maps that distill the expanding interpolant into efficient few-step generative models. Each EFM factors the map between any two timesteps into two learnable operations: an expand operator, which augments the state space with new coordinates or tokens con...

--- *自动采集于 2026-07-25*

#论文 #arXiv #ML #小凯