arXiv 2610.06801 · cs.CV · 2026-10-05
MC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion Transformers
Jiarui Chen, Zeqiang Lai, Jiangshan Wang, Ziheng Ouyang, Ye Huang, Xiangyu Yue, Cewu Lu, Chunchao Guo
Tencent Hunyuan
在最新榜 #179 本站收录于 2026-10-21
一句话要点
MC-Sparse 免训练地逐个挑选 KV token 并缓存补偿，缩小扩散 Transformer 稀疏与稠密注意力的质量差。
MC-Sparse is training-free sparse attention that selects individual KV tokens with cached compensation, closing the dense-sparse gap in DiTs.
阅读 arXiv 原文 PDF Hugging Face 讨论 代码仓库 项目页
主题与学科
同主题：视觉生成
- MaLiang-Harness: A Programmable Path to Image and Video Generation ▲367
- Any Time, Any Space: A World-State-Centric Framework for Agentic Video Generation ▲364
- DiffusionART: Efficient, Stable and Scalable Reinforcement of Diffusion Models ▲325
- UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation ▲197
- Matrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game Worlds ▲182
数据说明
元数据来自 arXiv API（已核实）；点赞来自 Hugging Face 每日论文；引用来自 Semantic Scholar。最近刷新 2026-10-22。
要点由本站根据摘要自写，不转载摘要原文，请阅读原文。