English
Related papers

Related papers: Rectified Sparse Attention

200 papers

Large language models (LLMs) have shown remarkable potential in processing long sequences and complex reasoning tasks, yet efficiently serving these models remains challenging due to the quadratic computational complexity of attention in…

Computation and Language · Computer Science 2025-04-22 Shang Yang , Junxian Guo , Haotian Tang , Qinghao Hu , Guangxuan Xiao , Jiaming Tang , Yujun Lin , Zhijian Liu , Yao Lu , Song Han

Modern large language models increasingly require long contexts for reasoning and multi-document tasks, but attention's quadratic complexity creates a severe computational bottleneck. We present Block-Sparse FlashAttention (BSFA), a drop-in…

Machine Learning · Computer Science 2025-12-09 Daniel Ohayon , Itay Lamprecht , Itay Hubara , Israel Cohen , Daniel Soudry , Noam Elata

Generating long sequences of tokens given a long-context input is a very compute-intensive inference scenario for large language models (LLMs). One prominent inference speed-up approach is to construct a smaller key-value (KV) cache,…

Computation and Language · Computer Science 2025-03-04 Fangyuan Xu , Tanya Goyal , Eunsol Choi

Efficient inference is critical for long-context language models, where attention computation and KV-cache access dominate the cost. Recent work RAT+, introduces a recurrence-augmented attention backbone that enables flexible dilated…

Machine Learning · Computer Science 2026-05-28 Xiuying Wei , Caglar Gulcehre

Recent advances in large language models highlighted the excessive quadratic cost of self-attention. Despite the significant research efforts, subquadratic attention methods still suffer from inferior performance in practice. We hypothesize…

Machine Learning · Computer Science 2025-05-02 Piotr Piękos , Róbert Csordás , Jürgen Schmidhuber

Programming-based Pre-trained Language Models (PPLMs) such as CodeBERT have achieved great success in many downstream code-related tasks. Since the memory and computational complexity of self-attention in the Transformer grow quadratically…

Computation and Language · Computer Science 2022-05-30 Tingting Liu , Chengyu Wang , Cen Chen , Ming Gao , Aoying Zhou

Scaling large language models to long contexts is challenging due to the quadratic computational cost of full attention. Mitigation approaches include KV-cache selection or compression techniques. We instead provide an effective and…

Machine Learning · Computer Science 2026-04-24 Yuzhen Mao , Michael Y. Li , Emily B. Fox

While diffusion language models (DLMs) offer a promising alternative to autoregressive models (ARs), existing open-source DLMs suffer from high inference latency. This bottleneck is mainly due to the attention's quadratic complexity with…

Computation and Language · Computer Science 2025-09-30 Zeqing Wang , Gongfan Fang , Xinyin Ma , Xingyi Yang , Xinchao Wang

Long-sequence processing is a critical capability for modern large language models. However, the self-attention mechanism in the standard Transformer architecture faces severe computational and memory bottlenecks when processing long…

Computation and Language · Computer Science 2025-09-30 Weilin Zhao , Zihan Zhou , Zhou Su , Chaojun Xiao , Yuxuan Li , Yanghao Li , Yudi Zhang , Weilun Zhao , Zhen Li , Yuxiang Huang , Ao Sun , Xu Han , Zhiyuan Liu

The Transformer architecture, despite its widespread success, struggles with long-context scenarios due to quadratic computation and linear memory growth. While various linear attention variants mitigate these efficiency constraints by…

Machine Learning · Computer Science 2025-10-02 Yuqi Pan , Yongqi An , Zheng Li , Yuhong Chou , Ruijie Zhu , Xiaohui Wang , Mingxuan Wang , Jinqiao Wang , Guoqi Li

Decoding throughput improvements from larger inference batches are limited by GPU memory, which is largely consumed by the key-value (KV) cache. Prior training-free KV cache offloading alleviates this by keeping redundant context on the CPU…

Computation and Language · Computer Science 2026-01-30 Yuxiang Huang , Pengjie Wang , Jicheng Han , Weilin Zhao , Zhou Su , Ao Sun , Hongya Lyu , Hengyu Zhao , Yudong Wang , Chaojun Xiao , Xu Han , Zhiyuan Liu

Attention is the dominant source of latency during long-context LLM inference, an increasingly popular workload with reasoning models and RAG. We propose Kascade, a training-free sparse attention method that leverages known observations…

Machine Learning · Computer Science 2025-12-19 Dhruv Deshmukh , Saurabh Goyal , Nipun Kwatra , Ramachandran Ramjee

Video diffusion Transformer (DiT) models excel in generative quality but hit major computational bottlenecks when producing high-resolution, long-duration videos. The quadratic complexity of full attention leads to prohibitively high…

Computer Vision and Pattern Recognition · Computer Science 2026-01-01 Chenlu Zhan , Wen Li , Chuyu Shen , Jun Zhang , Suhui Wu , Hao Zhang

The quadratic complexity of attention mechanisms poses a critical bottleneck for large language models processing long contexts. While dynamic sparse attention methods offer input-adaptive efficiency, they face fundamental trade-offs:…

Computation and Language · Computer Science 2026-02-06 Siran Liu , Guoxia Wang , Sa Wang , Jinle Zeng , HaoYang Xie , Siyu Lou , JiaBin Yang , DianHai Yu , Haifeng Wang , Chao Yang

Dynamic sparse attention (DSA) reduces the per-token attention bandwidth by restricting computation to a top-k subset of cached key-value (KV) entries, but its token-dependent selection pattern introduces a system-level challenge: the KV…

Hardware Architecture · Computer Science 2026-03-17 Noam Levy

Token-level sparse attention mechanisms, exemplified by DeepSeek Sparse Attention (DSA), achieve fine-grained key selection by scoring every historical key for each query through a lightweight indexer, then computing attention only on the…

In recent years, Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide array of text-centric tasks. However, their `large' scale introduces significant computational and storage challenges, particularly in…

Computation and Language · Computer Science 2024-07-03 Chaoran Zhang , Lixin Zou , Dan Luo , Min Tang , Xiangyang Luo , Zihao Li , Chenliang Li

We present core attention disaggregation (CAD), a technique that improves long-context large language model training by decoupling the core attention computation, softmax(QK^T)V, from the rest of the model and executing it on a separate…

Machine Learning · Computer Science 2025-10-22 Yonghao Zhuang , Junda Chen , Bo Pang , Yi Gu , Yibo Zhu , Yimin Jiang , Ion Stoica , Eric Xing , Hao Zhang

The transformer architecture has driven breakthroughs in recent years on tasks which require modeling pairwise relationships between sequential elements, as is the case in natural language understanding. However, long seqeuences pose a…

Computation and Language · Computer Science 2024-03-26 Heejun Lee , Jina Kim , Jeffrey Willette , Sung Ju Hwang

LLM decoding is bottlenecked for large batches and long contexts by loading the key-value (KV) cache from high-bandwidth memory, which inflates per-token latency, while the sequential nature of decoding limits parallelism. We analyze the…

Machine Learning · Computer Science 2025-05-28 Ted Zadouri , Hubert Strauss , Tri Dao