SCOPE: Subspace Clustering with Online Per-Head Top-K Estimation for Sparse Video Attention

• URL: http://arxiv.org/abs/2608.12780v1
• Date: Thu, 13 Aug 2026 03:43:31 GMT
• Title: SCOPE: Subspace Clustering with Online Per-Head Top-K Estimation for Sparse Video Attention
• Abstract summary: SCOPE is a training-free sparse attention framework for video-DiT inference.<n>It combines 3D-RoPE-aligned key subspace with clustering online per-head Top-$k$ estimation.<n>SCOPE consistently outperforms existing training-free baselines in both fidelity and latency.
• Score: 27.426483415833072
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Diffusion Transformers (DiTs) incur quadratic self-attention cost over spatiotemporal tokens. Existing training-free sparse attention methods often construct sparse masks from block-level or cluster-level proxy scores, which can obscure fine-grained differences among keys and miss high contribution keys under aggressive sparsity. Moreover, such proxy scores may yield overly concentrated softmax distributions, causing Top-$p$ to retain too few keys for some query clusters. Although a fixed Top-$k$ minimum alleviates this failure mode, a shared value cannot adapt to variations across heads and inputs. To address both limitations, we propose SCOPE, a training-free sparse attention framework that combines 3D-RoPE-aligned key subspace clustering with online per-head Top-$k$ estimation for efficient video-DiT inference. SCOPE partitions post-RoPE keys into temporal, height, and width subspaces, clusters them independently, and aggregates the corresponding centroid scores through lookup tables to obtain per key proxy scores for each query cluster. Building on existing hybrid Top-$p$/fixed Top-$k$ selection, SCOPE derives a head-specific Top-$k$ value online by averaging the initial retained key counts within each head, weighted by query cluster size, and selects additional keys only for query clusters whose initial retained key counts fall below this value. Sparse attention is then computed over the selected original keys and values. Across six model--task configurations, SCOPE consistently outperforms existing training-free baselines in both fidelity and latency, achieving up to a $1.99\times$ end-to-end speedup on 720p HunyuanVideo with $28.46$ dB PSNR relative to dense attention.

Related papers

• ClusterAttention: A training-free speedup of bidirectional attention [0.0]
  ClusterAttention is a training-free speedup of bidirectional attention layers.<n>On large-scale data ClusterAttention speeds up TabPFN-3 arXiv:2605.13986 by two to six times.
  arXiv  Detail & Related papers  (2026-08-27T11:04:41Z)
• DynaCalKV: Key-Value Cache Compression via Head Grouping and Adaptive Rank Allocation [0.0]
  We propose an improved low-rank KV cache compression framework.<n>We show that our method reduces the number of Key cache parameters while maintaining competitive accuracy.
  arXiv  Detail & Related papers  (2026-07-27T12:08:56Z)
• FASA: Frequency-aware Sparse Attention [56.26881872333624]
  We propose FASA, a novel framework that achieves query-aware token eviction by dynamically predicting token importance.<n>Our key finding is that a small, identifiable subset of "dominant" FCs consistently exhibits high contextual agreement with the full attention head.<n>Across a spectrum of long-context tasks, FASA consistently outperforms all token-eviction baselines and achieves near-oracle accuracy.
  arXiv  Detail & Related papers  (2026-02-03T06:09:06Z)
• Multipole Semantic Attention: A Fast Approximation of Softmax Attention for Pretraining [36.70261826975876]
  We present Multipole Semantic Attention (MuSe), an efficient approximation of softmax attention.<n>Our method addresses the quadratic computational complexity of transformers in the context length by clustering queries and keys separately.<n>For causal attention, we develop a hierarchical block decomposition that combines exact local computation with efficient long-range approximation.
  arXiv  Detail & Related papers  (2025-09-12T16:58:17Z)
• Squeezed Attention: Accelerating Long Context Length LLM Inference [61.787865959140994]
  We propose Squeezed Attention to accelerate applications where a large portion of the input context is fixed.<n>During inference, we compare query tokens from the user input with the centroids to predict which keys from the fixed context are semantically relevant.<n>We also present a hierarchical version of our algorithm which can reduce the complexity of attention from linear to logarithmic with respect to the fixed context length.
  arXiv  Detail & Related papers  (2024-11-14T18:54:19Z)
• Order-Optimal Regret in Distributed Kernel Bandits using Uniform Sampling with Shared Randomness [9.731329071569018]
  We consider distributed kernel bandits where $N$ agents aim to collaboratively maximize an unknown reward function. We develop the first algorithm that achieves the optimal regret order with a communication cost that is sublinear in both $N$ and $T$.
  arXiv  Detail & Related papers  (2024-02-20T17:49:10Z)
• Revisiting Foreground and Background Separation in Weakly-supervised Temporal Action Localization: A Clustering-based Approach [48.684550829098534]
  Weakly-supervised temporal action localization aims to localize action instances in videos with only video-level action labels. We propose a novel clustering-based F&B separation algorithm. We evaluate our method on three benchmarks: THUMOS14, ActivityNet v1.2 and v1.3.
  arXiv  Detail & Related papers  (2023-12-21T18:57:12Z)
• Hard Regularization to Prevent Deep Online Clustering Collapse without Data Augmentation [65.268245109828]
  Online deep clustering refers to the joint use of a feature extraction network and a clustering model to assign cluster labels to each new data point or batch as it is processed. While faster and more versatile than offline methods, online clustering can easily reach the collapsed solution where the encoder maps all inputs to the same point and all are put into a single cluster. We propose a method that does not require data augmentation, and that, differently from existing methods, regularizes the hard assignments.
  arXiv  Detail & Related papers  (2023-03-29T08:23:26Z)
• DeepCluE: Enhanced Image Clustering via Multi-layer Ensembles in Deep Neural Networks [53.88811980967342]
  This paper presents a Deep Clustering via Ensembles (DeepCluE) approach. It bridges the gap between deep clustering and ensemble clustering by harnessing the power of multiple layers in deep neural networks. Experimental results on six image datasets confirm the advantages of DeepCluE over the state-of-the-art deep clustering approaches.
  arXiv  Detail & Related papers  (2022-06-01T09:51:38Z)
• Overcomplete Deep Subspace Clustering Networks [80.16644725886968]
  Experimental results on four benchmark datasets show the effectiveness of the proposed method over DSC and other clustering methods in terms of clustering error. Our method is also not as dependent as DSC is on where pre-training should be stopped to get the best performance and is also more robust to noise.
  arXiv  Detail & Related papers  (2020-11-16T22:07:18Z)
• Differentiable Hierarchical Graph Grouping for Multi-Person Pose Estimation [95.72606536493548]
  Multi-person pose estimation is challenging because it localizes body keypoints for multiple persons simultaneously. We propose a novel differentiable Hierarchical Graph Grouping (HGG) method to learn the graph grouping in bottom-up multi-person pose estimation task.
  arXiv  Detail & Related papers  (2020-07-23T08:46:22Z)
• Fast Transformers with Clustered Attention [14.448898156256478]
  We propose clustered attention, which instead of computing the attention for every query, groups queries into clusters and computes attention just for the centroids. This results in a model with linear complexity with respect to the sequence length for a fixed number of clusters. We evaluate our approach on two automatic speech recognition datasets and show that our model consistently outperforms vanilla transformers for a given computational budget.
  arXiv  Detail & Related papers  (2020-07-09T14:17:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.