STARFlow-V：基于自回归归一化流的端到端视频生成建模
STARFlow-V: End-to-End Video Generative Modeling with Autoregressive Normalizing Flows
为视觉研究者提供了一条绕开扩散模型的视频生成路线：可逆归一化流在时空潜空间中做自回归因果预测，天然支持似然估计与 T2V/I2V/V2V 统一建模，其 flow-score matching 与视频感知 Jacobi 迭代可直接迁移到其他自回归视频生成器的误差累积与采样加速问题上。
原文依据 STARFlow-V, a normalizing flow-based video generator with substantial benefits such as end-to-end learning, robust causal prediction and native likelihood estimation
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论归一化流（NF）是针对连续数据的端到端、基于似然的生成模型，近期在图像生成上的进展使其重新受到关注。然而在时空复杂度与计算开销都显著更高的视频生成领域，最优系统几乎完全依赖扩散模型。本文重新审视这一设计空间，提出 STARFlow-V——一个基于归一化流的视频生成器，具备端到端学习、稳健的因果预测与原生似然估计等重要优势。STARFlow-V 建立在近期提出的 STARFlow 之上，在时空潜空间中运行，采用全局-局部架构：将因果依赖限制在全局潜空间中，同时保留丰富的帧内局部交互。这缓解了随时间推移的误差累积——这是标准自回归扩散模型生成的一个常见缺陷。此外，我们提出流-分数匹配（flow-score matching），为模型配备一个轻量级因果去噪器，以自回归方式提升视频生成的一致性。为提升采样效率，STARFlow-V 采用视频感知的 Jacobi 迭代方案，将内部更新重构为可并行化的迭代而不破坏因果性。得益于可逆结构，同一模型可原生支持文本到视频、图像到视频以及视频到视频生成任务。实验上，STARFlow-V 相对于基于扩散的基线取得了较强的视觉保真度与时序一致性，并具有实用的采样吞吐量。据我们所知，这些结果首次证明归一化流能够实现高质量的自回归视频生成，从而将其确立为构建世界模型的一个有前景的研究方向。
英文原文摘要 1824 字符
Normalizing flows (NFs) are end-to-end likelihood-based generative models for continuous data, and have recently regained attention with encouraging progress on image generation. Yet in the video generation domain, where spatiotemporal complexity and computational cost are substantially higher, state-of-the-art systems almost exclusively rely on diffusion-based models. In this work, we revisit this design space by presenting STARFlow-V, a normalizing flow-based video generator with substantial benefits such as end-to-end learning, robust causal prediction and native likelihood estimation. Building upon the recently proposed STARFlow, STARFlow-V operates in the spatiotemporal latent space with a global-local architecture which restricts causal dependencies to a global latent space while preserving rich local within-frame interactions. This eases error accumulation over time, a common pitfall of standard autoregressive diffusion model generation. Additionally, we propose flow-score matching, which equips the model with a light-weight causal denoiser to improve the video generation consistency in an autoregressive fashion. To improve the sampling efficiency, STARFlow-V employs a video-aware Jacobi iteration scheme that recasts inner updates as parallelizable iterations without breaking causality. Thanks to the invertible structure, the same model can natively support text-to-video, image-to-video as well as video-to-video generation tasks. Empirically, STARFlow-V achieves strong visual fidelity and temporal consistency with practical sampling throughput relative to diffusion-based baselines. These results present the first evidence, to our knowledge, that NFs are capable of high-quality autoregressive video generation, establishing them as a promising research direction for building world models.