arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.10099cs.LGcs.CV
收藏

流形上的学习:通过表示编码器解锁标准扩散变压器

Learning on the Manifold: Unlocking Standard Diffusion Transformers with Representation Encoders

Johns Hopkins University(约翰霍普金斯大学)

Amandeep Kumar, Vishal M. Patel

AI总结:

研究利用表示编码器进行生成建模时标准扩散变压器收敛失败的问题,指出根本原因是几何干扰,提出黎曼流匹配与雅可比正则化方法,使标准扩散变压器无需宽度缩放就能收敛。

摘要:

利用表示编码器进行生成建模为高效、高保真合成提供了一条途径。然而,标准扩散变压器无法直接在这些表示上收敛。虽然最近的工作将此归因于容量瓶颈并提出了计算成本高昂的扩散变压器宽度缩放方法,但我们证明失败从根本上说是几何问题。我们将几何干扰识别为根本原因:标准欧几里得流匹配迫使概率路径穿过表示编码器超球面特征空间的低密度内部,而不是沿着流形表面。为了解决这个问题,我们提出了带有雅可比正则化的黎曼流匹配(RJF)。通过将生成过程约束到流形测地线并校正曲率引起的误差传播,RJF使标准扩散变压器架构能够在不进行宽度缩放的情况下收敛。我们的方法RJF使标准DiT - B架构(131M参数)能够有效收敛,在先前方法无法收敛的情况下实现了3.37的FID。代码:此https URL

英文摘要:

Leveraging representation encoders for generative modeling offers a path for efficient, high-fidelity synthesis. However, standard diffusion transformers fail to converge on these representations directly. While recent work attributes this to a capacity bottleneck proposing computationally expensive width scaling of diffusion transformers we demonstrate that the failure is fundamentally geometric. We identify Geometric Interference as the root cause: standard Euclidean flow matching forces probability paths through the low-density interior of the hyperspherical feature space of representation encoders, rather than following the manifold surface. To resolve this, we propose Riemannian Flow Matching with Jacobi Regularization (RJF). By constraining the generative process to the manifold geodesics and correcting for curvature-induced error propagation, RJF enables standard Diffusion Transformer architectures to converge without width scaling. Our method RJF enables the standard DiT-B architecture (131M parameters) to converge effectively, achieving an FID of 3.37 where prior methods fail to converge. Code: https://github.com/amandpkr/RJF

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。

↑