English
Related papers

Related papers: Beyond Homogeneous Attention: Memory-Efficient LLM…

200 papers

Large language models are increasingly applied to multi-document and long-form inputs, yet long-context inference remains memory- and noise-inefficient. Key-value (KV) caching scales linearly with context length, while external retrieval…

Computation and Language · Computer Science 2026-01-29 Qingsen Ma , Dianyun Wang , Yaoye Wang , Lechen Ning , Sujie Zhu , Xiaohang Zhang , Jiaming Lyu , Linhao Ren , Zhenbo Xu , Zhaofeng He

Large language models (LLMs) rely on key-value (KV) caches for efficient autoregressive decoding; however, cache size grows linearly with context length and model depth, becoming a major bottleneck in long-context inference. Prior KV cache…

Machine Learning · Computer Science 2025-09-22 Dmitry Akulov , Mohamed Sana , Antonio De Domenico , Tareq Si Salem , Nicola Piovesan , Fadhel Ayed

Transformer-based architectures have become the prevailing backbone of large language models. However, the quadratic time and memory complexity of self-attention remains a fundamental obstacle to efficient long-context modeling. To address…

Computation and Language · Computer Science 2026-02-10 Yutao Sun , Zhenyu Li , Yike Zhang , Tengyu Pan , Bowen Dong , Yuyi Guo , Jianyong Wang

Large Language Models (LLMs) face limitations due to the high demand on GPU memory and computational resources when handling long contexts. While sparsify the Key-Value (KV) cache of transformer model is a typical strategy to alleviate…

Machine Learning · Computer Science 2024-02-06 Yumeng Wang , Zhenyang Xiao

While Transformer self-attention offers strong parallelism, the Key-Value (KV) cache grows linearly with sequence length and becomes a bottleneck for inference efficiency. Multi-head latent attention was recently developed to compress the…

Machine Learning · Computer Science 2025-11-04 Keqi Deng , Philip C. Woodland

Large language models (LLMs) now support extremely long context windows, but the quadratic complexity of vanilla attention results in significantly long Time-to-First-Token (TTFT) latency. Existing approaches to address this complexity…

Computation and Language · Computer Science 2025-09-04 Qianchao Zhu , Jiangfei Duan , Chang Chen , Siran Liu , Guanyu Feng , Xin Lv , Xiao Chuanfu , Dahua Lin , Chao Yang

Transformers, driven by attention mechanisms, form the foundation of large language models (LLMs). As these models scale up, efficient GPU attention kernels become essential for high-throughput and low-latency inference. Diverse LLM…

Distributed, Parallel, and Cluster Computing · Computer Science 2025-04-23 Zihao Ye , Lequn Chen , Ruihang Lai , Wuwei Lin , Yineng Zhang , Stephanie Wang , Tianqi Chen , Baris Kasikci , Vinod Grover , Arvind Krishnamurthy , Luis Ceze

Autoregressive Transformers rely on Key-Value (KV) caching to accelerate inference. However, the linear growth of the KV cache with context length leads to excessive memory consumption and bandwidth constraints. This bottleneck is…

Computation and Language · Computer Science 2025-06-10 Ravi Ghadia , Avinash Kumar , Gaurav Jain , Prashant Nair , Poulami Das

Current hierarchical attention methods, such as NSA and InfLLMv2, select the top-k relevant key-value (KV) blocks based on coarse attention scores and subsequently apply fine-grained softmax attention on the selected tokens. However, the…

Computation and Language · Computer Science 2026-05-19 Yuxiang Huang , Nuno M. T. Gonçalves , Federico Alvetreti , Lei Li , Xu Han , Edoardo M. Ponti , André F. T. Martins , Marcos V. Treviso

FlashAttention (Dao, 2023) effectively reduces the quadratic peak memory usage to linear in training transformer-based large language models (LLMs) on a single GPU. In this paper, we introduce DISTFLASHATTN, a distributed memory-efficient…

Machine Learning · Computer Science 2024-04-02 Dacheng Li , Rulin Shao , Anze Xie , Eric P. Xing , Xuezhe Ma , Ion Stoica , Joseph E. Gonzalez , Hao Zhang

Visual token reduction lowers inference costs caused by extensive image features in large vision-language models (LVLMs). Unlike relevant studies that prune tokens in self-attention-only LVLMs, our work uniquely addresses…

Computer Vision and Pattern Recognition · Computer Science 2025-04-02 Jewon Lee , Ki-Ung Song , Seungmin Yang , Donguk Lim , Jaeyeon Kim , Wooksu Shin , Bo-Kyeong Kim , Yong Jae Lee , Tae-Ho Kim

Large Language Models (LLMs) suffer inference-time memory bottlenecks dominated by the attention Key-Value (KV) cache, which scales with model size and context length. While KV-cache quantization alleviates this cost, bit allocation between…

Machine Learning · Computer Science 2026-05-12 Mohsen Hariri , Alan Luo , Weicong Chen , Shaochen Zhong , Tianyi Zhang , Qifan Wang , Xia Hu , Xiaotian Han , Vipin Chaudhary

Transformer-based language models utilize the attention mechanism for substantial performance improvements in almost all natural language processing (NLP) tasks. Similar attention structures are also extensively studied in several other…

Computation and Language · Computer Science 2023-05-17 Nurullah Sevim , Ege Ozan Özyedek , Furkan Şahinuç , Aykut Koç

The quadratic complexity of the attention module makes it gradually become the bulk of compute in Transformer-based LLMs during generation. Moreover, the excessive key-value cache that arises when dealing with long inputs also brings severe…

Computation and Language · Computer Science 2023-10-17 Siyu Ren , Qi Jia , Kenny Q. Zhu

Despite the significant success of large language models (LLMs), their extensive memory requirements pose challenges for deploying them in long-context token generation. The substantial memory footprint of LLM decoders arises from the…

Machine Learning · Computer Science 2024-02-12 Amir Zandieh , Insu Han , Vahab Mirrokni , Amin Karbasi

We present a novel non attention based architecture for large language models (LLMs) that efficiently handles very long context windows, on the order of hundreds of thousands to potentially millions of tokens. Unlike traditional Transformer…

Machine Learning · Computer Science 2025-06-04 Andrew Kiruluta , Preethi Raju , Priscilla Burity

Long-context decoding in LLMs is IO-bound: each token re-reads an ever-growing KV cache. Prior accelerations cut bytes via compression, which lowers fidelity, or selection/eviction, which restricts what remains accessible, and both can…

Machine Learning · Computer Science 2026-04-02 Jinghan Yao , Sam Adé Jacobs , Walid Krichene , Masahiro Tanaka , Dhabaleswar K Panda

Language models struggle to generalize beyond pretraining context lengths, limiting long-horizon reasoning and retrieval. Continued pretraining on long-context data can help but is expensive due to the quadratic scaling of Attention. We…

Computation and Language · Computer Science 2026-03-19 Sakshi Choudhary , Aditya Chattopadhyay , Luca Zancato , Elvis Nunez , Matthew Trager , Wei Xia , Stefano Soatto

Large language models (LLMs) face significant challenges in processing long contexts due to the linear growth of the key-value (KV) cache and quadratic complexity of self-attention. Existing approaches address these bottlenecks separately:…

Computation and Language · Computer Science 2026-04-17 Zeng You , Yaofo Chen , Qiuwu Chen , Ying Sun , Shuhai Zhang , Yingjian Li , Yaowei Wang , Mingkui Tan

The quadratic complexity of self-attention in Transformers has hindered the processing of long text. To alleviate this problem, previous works have proposed to sparsify the attention matrix, taking advantage of the observation that crucial…

Computation and Language · Computer Science 2024-01-12 Ziwei He , Jian Yuan , Le Zhou , Jingwen Leng , Bo Jiang