Scaling Diffusion Transformers to generate high-resolution, long videos is constrained by the quadratic cost of self-attention, and existing sparse attention methods degrade under high sparsity. We show empirically that generation quality is determined not by the sparsity ratio itself, but by how well the sparse mask aligns with the tile-wise geometry of full attention. Based on this insight, we propose Veda, a distilled sparse attention framework that formulates tile selection as an explicit reconstruction problem from full attention. Veda integrates statistics-aware tile scoring with head-aware tiling to reduce estimation error and structural mismatch, enabling aggressive sparsity. A hardware-efficient tile-skipping kernel converts theoretical sparsity into practical wall-clock speedups. Experiments on large video diffusion models, including Waver and Wan2.1, demonstrate substantial acceleration with no noticeable degradation in generation quality. To generate 720P 10-second videos on Waver-T2V-12B, Veda achieves a 5.1$\times$ end-to-end speedup and a 10.5$\times$ self-attention speedup, reducing attention overhead from 92% to 50%. Notably, the gains increase with sequence length, indicating that Veda scales favorably with spatiotemporal resolution across models.

翻译：缩放扩散Transformer以生成高分辨率、长时长视频受限于自注意力的二次复杂度，现有稀疏注意力方法在高稀疏度下会退化。我们通过实验证明，生成质量并非由稀疏比率本身决定，而是取决于稀疏掩码与全注意力的分块几何结构对齐程度。基于这一发现，我们提出Veda——一种蒸馏稀疏注意力框架，将分块选择建模为从全注意力恢复的显式重建问题。Veda融合了统计感知的分块评分与头感知分块策略，以降低估计误差和结构不匹配，从而实现高稀疏度。硬件高效的分块跳过内核可将理论稀疏度转化为实际时钟速度提升。在包括Waver和Wan2.1在内的大型视频扩散模型上的实验表明，该方法在生成质量无显著退化的情况下实现了大幅加速。针对Waver-T2V-12B生成720P 10秒视频，Veda实现了5.1倍端到端加速和10.5倍自注意力加速，将注意力开销从92%降至50%。值得注意的是，加速增益随序列长度增加，表明Veda在跨模型时空分辨率下具有可扩展性优势。