English
Related papers

Related papers: Kascade: A Practical Sparse Attention Method for L…

200 papers

Multimodal Large Language Models (MLLMs) are commonly derived by extending pre-trained Large Language Models (LLMs) with visual capabilities. In this work, we investigate how MLLMs process visual inputs by analyzing their attention…

Computer Vision and Pattern Recognition · Computer Science 2025-07-08 Jiahui Wang , Zuyan Liu , Yongming Rao , Jiwen Lu

Long contexts improve capabilities of large language models but pose serious hardware challenges: compute and memory footprints grow linearly with sequence length. Particularly, the decoding phase continuously accesses massive KV cache,…

Hardware Architecture · Computer Science 2026-04-29 Wang Fan , Wei Cao , Xi Zha , Kedi Ma , MingQian Sun , Jialin Chen , Fengzhe Zhang , Fan Zhang

Large language models face significant cost challenges in long-sequence inference. To address this, reusing historical Key-Value (KV) Cache for improved inference efficiency has become a mainstream approach. Recent advances further enhance…

Machine Learning · Computer Science 2025-08-19 Ziyi Cao , Qingyi Si , Jingbin Zhang , Bingquan Liu

Transformer-based large language models face severe scalability challenges in long-context generation due to the computational and memory costs of full-context attention. Under practical computation and memory constraints, many…

Computation and Language · Computer Science 2026-05-13 Xianpeng Shang , Jiang Li , Zehua Duo , Qianyi Cai , Xiangdong Su

Long-context inference in large language models (LLMs) is increasingly constrained by the KV cache bottleneck: memory usage grows linearly with sequence length, while attention computation scales quadratically. Existing approaches address…

Computation and Language · Computer Science 2025-11-13 Huanxuan Liao , Yixing Xu , Shizhu He , Guanchen Li , Xuanwu Yin , Dong Li , Emad Barsoum , Jun Zhao , Kang Liu

Diffusion Transformers (DiTs) set the state of the art in visual generation, yet their quadratic self-attention cost fundamentally limits scaling to long token sequences. Recent Top-K sparse attention approaches reduce the computation of…

Computer Vision and Pattern Recognition · Computer Science 2025-12-19 Yifan Zhou , Zeqi Xiao , Tianyi Wei , Shuai Yang , Xingang Pan

Autoregressive decoding with generative Large Language Models (LLMs) on accelerators (GPUs/TPUs) is often memory-bound where most of the time is spent on transferring model parameters from high bandwidth memory (HBM) to cache. On the other…

Machine Learning · Computer Science 2024-02-15 Yashas Samaga B L , Varun Yerram , Chong You , Srinadh Bhojanapalli , Sanjiv Kumar , Prateek Jain , Praneeth Netrapalli

LLMs encounter significant challenges in resource consumption nowadays, especially with long contexts. Despite extensive efforts dedicate to enhancing inference efficiency, these methods primarily exploit internal sparsity within the…

Computation and Language · Computer Science 2025-07-17 Yi Zhao , Zuchao Li , Hai Zhao

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

Long-context inference increasingly operates over CPU-resident KV caches, either because decoding-time KV states exceed GPU memory capacity or because disaggregated prefill-decode systems place KV data in host memory. Although block-sparse…

Machine Learning · Computer Science 2026-05-11 Feiyu Yao , Zhixiong Niu , Xiaqing Li , Yongqiang Xiong , Juan Fang , Qian Wang

Sparse attention mechanisms aim to reduce computational overhead with minimal accuracy loss by selectively processing salient tokens. Despite their effectiveness, most methods merely exploit a model's inherent sparsity and thus plateau at…

Machine Learning · Computer Science 2026-03-02 Feng Chen , Yefei He , Lequan Lin , Chenhui Gou , Jing Liu , Bohan Zhuang , Qi Wu

Multi-head Latent Attention (MLA) is an innovative architecture proposed by DeepSeek, designed to ensure efficient and economical inference by significantly compressing the Key-Value (KV) cache into a latent vector. Compared to MLA,…

Computation and Language · Computer Science 2025-10-06 Tao Ji , Bin Guo , Yuanbin Wu , Qipeng Guo , Lixing Shen , Zhan Chen , Xipeng Qiu , Qi Zhang , Tao Gui

Large Language Models (LLMs) exhibit enhanced capabilities by Chain-of-Thought reasoning. However, the extended reasoning sequences introduce significant GPU memory overhead due to increased key-value (KV) cache. Existing KV cache…

Machine Learning · Computer Science 2025-10-16 Haoyue Zhang , Hualei Zhang , Xiaosong Ma , Jie Zhang , Song Guo

With the widespread deployment of long-context large language models (LLMs), there has been a growing demand for efficient support of high-throughput inference. However, as the key-value (KV) cache expands with the sequence length, the…

Machine Learning · Computer Science 2025-04-29 Hanshi Sun , Li-Wen Chang , Wenlei Bao , Size Zheng , Ningxin Zheng , Xin Liu , Harry Dong , Yuejie Chi , Beidi Chen

Large Language Models capable of handling extended contexts are in high demand, yet their inference remains challenging due to substantial Key-Value cache size and high memory bandwidth requirements. Previous research has demonstrated that…

Machine Learning · Computer Science 2025-10-29 Junlin Mu , Hantao Huang , Jihang Zhang , Minghui Yu , Tao Wang , Yidong Li

In this work, we propose an extreme compression technique for Large Multimodal Models (LMMs). While previous studies have explored quantization as an efficient post-training compression method for Large Language Models (LLMs), low-bit…

Computer Vision and Pattern Recognition · Computer Science 2025-03-11 Mohsen Gholami , Mohammad Akbari , Kevin Cannons , Yong Zhang

KV-cache retrieval is essential for long-context LLM inference, yet existing methods struggle with distribution drift and high latency at scale. We introduce ParisKV, a drift-robust, GPU-native KV-cache retrieval framework based on…

Machine Learning · Computer Science 2026-02-11 Yanlin Qi , Xinhang Chen , Huiqiang Jiang , Qitong Wang , Botao Peng , Themis Palpanas

We present Prompt Cache, an approach for accelerating inference for large language models (LLM) by reusing attention states across different LLM prompts. Many input prompts have overlapping text segments, such as system messages, prompt…

Computation and Language · Computer Science 2024-04-26 In Gim , Guojun Chen , Seung-seob Lee , Nikhil Sarda , Anurag Khandelwal , Lin Zhong

Attention efficiency is critical to large language model (LLM) inference. While prior advances optimize attention execution for individual requests (e.g., FlashAttention), production LLM serving relies on batching requests with highly…

Distributed, Parallel, and Cluster Computing · Computer Science 2026-02-09 Rui Ning , Wei Zhang , Fan Lai

Efficient inference on GPUs using large language models remains challenging due to memory bandwidth limitations, particularly during data transfers between High Bandwidth Memory (HBM) and SRAM in attention computations. Approximate…

Machine Learning · Computer Science 2025-06-06 Nirav Koley , Prajwal Singhania , Abhinav Bhatele