当数字发声：文本到视频扩散模型中文本数词与视觉实例的对齐
When Numbers Speak: Aligning Textual Numerals and Visual Instances in Text-to-Video Diffusion Models
通过筛选判别性自注意力/交叉注意力头提取可计数的潜变量布局，再调制交叉注意力引导重生成，是一种无需训练即可复用的生成过程结构引导手段，可直接迁移到其他需要空间/数量约束的扩散生成任务。
原文依据 NUMINA identifies prompt-layout inconsistencies by selecting discriminative self- and cross-attention heads to derive a countable latent layout.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论文本到视频扩散模型已实现开放式视频合成，但往往难以生成提示词中所指定数量的物体。我们提出 NUMINA，一个无需训练的「先识别、后引导」框架，用于改善数值对齐。NUMINA 通过选取具有判别性的自注意力与交叉注意力头，推导出可计数的潜变量布局，从而识别提示词与布局之间的不一致。随后它保守地精化该布局，并调制交叉注意力以引导重新生成。在本文提出的 CountBench 上，NUMINA 在 Wan2.1-1.3B 上将计数准确率最高提升 7.4%，在 5B 与 14B 模型上分别提升 4.9% 与 5.5%。此外，CLIP 对齐得到改善，同时保持了时序一致性。这些结果表明，结构引导可与种子搜索和提示词增强形成互补，为计数准确的文本到视频扩散提供了一条实用路径。代码已发布于 https://github.com/H-EmbodVis/NUMINA。
英文原文摘要 967 字符
Text-to-video diffusion models have enabled open-ended video synthesis, but often struggle with generating the correct number of objects specified in a prompt. We introduce NUMINA, a training-free identify-then-guide framework for improved numerical alignment. NUMINA identifies prompt-layout inconsistencies by selecting discriminative self- and cross-attention heads to derive a countable latent layout. It then refines this layout conservatively and modulates cross-attention to guide regeneration. On the introduced CountBench, NUMINA improves counting accuracy by up to 7.4% on Wan2.1-1.3B, and by 4.9% and 5.5% on 5B and 14B models, respectively. Furthermore, CLIP alignment is improved while maintaining temporal consistency. These results demonstrate that structural guidance complements seed search and prompt enhancement, offering a practical path toward count-accurate text-to-video diffusion. The code is available at https://github.com/H-EmbodVis/NUMINA.