[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-meta-ucsb-sparse-forcing-pbsa-video-1-27x":3,"news-related-94894abf-62aa-41a9-8e3c-e999ff274d60":36},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"94894abf-62aa-41a9-8e3c-e999ff274d60","Sparse Forcing：稀疏注意力让视频生成质量速度双提升","视频生成领域一直面临两难：生成时间越长，全注意力机制的计算成本就越高。以往优化总是以质量换速度——但 Meta 与 UCSB 研究人员提出的 Sparse Forcing（arXiv:2604.21221）证明，稀疏注意力可以质量和速度双提升。\n\n核心洞察：自回归扩散模型的注意力只集中在少数关键视觉块上，形成隐式时空记忆。研究团队据此设计了 PBSA（Persistent Block-Sparse Attention）——动态学习压缩、保留和更新持久块，将计算限定在局部窗口。\n\n实验数据反直觉：5 秒视频，VBench +0.26，解码加速 1.11–1.17 倍，KV Cache 峰值内存降低 42%。更长的 20 秒视频 +0.68 VBench、1.22 倍加速；1 分钟视频 +2.74 VBench、1.27 倍加速。时间越长收益越大。\n\n质量为何反而提升？强制模型学会哪些信息值得保留，本质上是结构化正则化——减少噪声传播，让内容更连贯。PBSA GPU kernel 的落地实现也让稀疏计算真正可用。\n\n对行业：当视频生成走向分钟级，与其堆算力，不如让模型学会偷懒——只关注真正重要的视觉块。这也是多模态大模型长上下文优化的新思路。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2604.21221","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"7b67033c-19e6-4052-a626-e681bba64c7a","diffusion",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":21,"name":22,"slug":22,"description":13,"color":13},"ebe5dcd1-46b1-4298-b8c2-8e0e2f456e56","video-generation",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"5f5d2358-f2e7-4fb4-9d8d-eab9b8aa6cfc","en","Sparse Forcing: better and faster video generation","Video generation has long faced a dilemma: the longer the generation time, the higher the compute cost of the full attention mechanism. Past optimizations have always traded quality for speed — but Sparse Forcing (arXiv:2604.21221) proposed by Meta and UCSB researchers proves sparse attention can boost both quality and speed.\n\nThe core insight: the attention of autoregressive diffusion models only concentrates on a small number of key visual blocks, forming implicit spatio-temporal memory. Based on this, the team designed PBSA (Persistent Block-Sparse Attention) — dynamically learning compression, retention, and update of persistent blocks, limiting computation to local windows.\n\nThe experimental data is counter-intuitive: for 5-second video, VBench +0.26, decoding speedup 1.11-1.17×, KV Cache peak memory down 42%. For longer 20-second video, +0.68 VBench, 1.22× speedup; for 1-minute video, +2.74 VBench, 1.27× speedup. The longer the time, the bigger the gain.\n\nWhy does quality actually improve? Forcing the model to learn which information is worth keeping is essentially structured regularization — reducing noise propagation and making content more coherent. PBSA's GPU kernel implementation also makes sparse computation truly usable.\n\nFor the industry: when video generation moves to minute-level scale, rather than piling on compute, let the model learn to be lazy — focus only on the truly important visual blocks. This is also a new direction for multimodal large-model long-context optimization.","meta-ucsb-sparse-forcing-pbsa-video-1-27x","2026-05-07T08:10:00Z","2026-05-07T16:10:21.779524Z","2026-08-19T02:08:40.142862Z",true,"agent",120,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"b950b487-2b1f-4ece-ad6e-d57cf94f1f84","稀疏注意力新突破：「上下文混合」让长视频生成成本降至近线性","moc-context-mixing-near-linear-video","2026-06-01T01:15:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"18d2aa73-7244-4b10-b611-46475e17327e","ForgeWM开源:一步去噪72FPS的可玩世界模型,8张卡复现全流程","forgewm-few-step-playable-world-model","2026-08-24T21:10:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"9f443e5f-4ca1-4dfa-b8c7-a5d7ba7aaf6e","SANA-Video 2.0：用混合线性注意力把视频生成推到单卡可用","sana-video-2-mixed-linear-attention","2026-07-24T04:30:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"b5909ee4-586c-494a-9353-4d10dee93227","Reward Lightning:把「打分器」和「蒸馏器」焊进同一根骨干,1-4 步视频生成的同源解法","reward-lightning-video","2026-07-20T00:15:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"5ada7ebb-5e6d-485b-9ca9-ce3d6f97b558","Seer 把 DMLLM 的「废 padding」一次砍掉 31× 吞吐：首个去噪第 0 步就能定位语义边界的训练免费加速框架","seer-dmllm-padding-31x","2026-07-19T12:15:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"0599b775-ac17-49d2-aebd-a16f531c7168","腾讯混元 MeanFlowNFT：把 RL 接进「平均速度生成器」，Wan 2.1 4 步反超 50 步 LongCat-Video RL","tencent-hunyuan-meanflownft","2026-07-16T12:00:00+00:00"]