When Geometry Aligns: Dihedral Hidden-State Transformations in UNet, ViT, and DiT Architectures - 深度解析 论文来源:ArXiv (2607.03580) 作者:Mojtaba Faramarzi, Alex Lamb, Irina Rish 分类:cs.LG, cs.CV 发布时间:2026-07-03T19:47:07Z 解读时间:2026年07月07日 09:02:04 📋 论文基本信息 标题:When Geometry Aligns: Dihedral Hidden-State Transformations in UNet, ViT, and DiT Architectures
论文来源:ArXiv (2607.03580)
作者:Mojtaba Faramarzi, Alex Lamb, Irina Rish
分类:cs.LG, cs.CV
发布时间:2026-07-03T19:47:07Z
解读时间:2026年07月07日 09:02:04
标题:When Geometry Aligns: Dihedral Hidden-State Transformations in UNet, ViT, and DiT Architectures
作者:Mojtaba Faramarzi, Alex Lamb, Irina Rish
ArXiv ID:2607.03580
链接:https://arxiv.org/abs/2607.03580v1
分类:cs.LG, cs.CV
研究领域:多模态
本论文研究了 多模态 领域的重要问题。
Diffusion architectures now encompass convolutional UNets as well as transformer-based designs such as Diffusion Transformers (DiTs), inspired by Vision Transformers (ViTs), yet the effects of structured geometric perturbations within these architectures remain poorly understood. We study this question through a unified framework that applies reflection-based elements of the dihedral group to intermediate hidden states as controlled internal interventions, contrasting geometrically consistent and inconsistent variants. Using activation-level diagnostics, including Self-Consistency Shift (SCS), Activation Mass Scatter (AMS), and Drift, we analyze feature stability and geometric drift. We find that consistent transformations improve stability, while inconsistent ones induce predictable, arch
该研究对于解决当前领域面临的挑战具有重要意义。
论文提出了一种新颖的方法来解决相关问题。
论文通过大量实验验证了所提方法的有效性。
本论文的主要创新点包括:
该方法在 多模态 领域具有广阔的应用前景。
建议读者根据自身需求深入阅读相关文献。
本论文为相关研究做出了重要贡献。
本文由 AI 自动生成。要启用 Qwen 深度分析,请配置 DASHSCOPE_API_KEY。