English
Related papers

Related papers: Linearized Attention Cannot Enter the Kernel Regim…

200 papers

Linear attention significantly reduces the computational complexity of Transformers from quadratic to linear, yet it consistently lags behind softmax-based attention in performance. We identify the root cause of this degradation as the…

Machine Learning · Computer Science 2026-02-05 Weikang Meng , Liangyu Huo , Yadan Luo , Yaowei Wang , Yingjian Li , Zheng Zhang

The transformer is the most popular neural architecture for language modeling. The cornerstone of the transformer is its global attention mechanism, which lets the model aggregate information from all preceding tokens before generating the…

Computation and Language · Computer Science 2026-05-20 Jiaoda Li , Ryan Cotterell

An attention matrix of a transformer self-attention sublayer can provably be decomposed into two components and only one of them (effective attention) contributes to the model output. This leads us to ask whether visualizing effective…

Computation and Language · Computer Science 2021-05-20 Kaiser Sun , Ana Marasović

Transformers have achieved extraordinary success in modern machine learning due to their excellent ability to handle sequential data, especially in next-token prediction (NTP) tasks. However, the theoretical understanding of their…

Machine Learning · Computer Science 2024-10-01 Ruiquan Huang , Yingbin Liang , Jing Yang

Understanding the reasons behind the exceptional success of transformers requires a better analysis of why attention layers are suitable for NLP tasks. In particular, such tasks require predictive models to capture contextual meaning which…

Machine Learning · Statistics 2024-05-20 Simone Bombari , Marco Mondelli

Recent theoretical works based on the neural tangent kernel (NTK) have shed light on the optimization and generalization of over-parameterized networks, and partially bridge the gap between their practical success and classical learning…

Machine Learning · Computer Science 2020-08-10 Kyung-Su Kim , Aurélie C. Lozano , Eunho Yang

We perform a study on the generalization ability of the wide two-layer ReLU neural network on $\mathbb{R}$. We first establish some spectral properties of the neural tangent kernel (NTK): $a)$ $K_{d}$, the NTK defined on $\mathbb{R}^{d}$,…

Machine Learning · Statistics 2023-02-14 Jianfa Lai , Manyun Xu , Rui Chen , Qian Lin

Linear attention reduces the quadratic cost of softmax attention to $\mathcal{O}(T)$, but its memory state grows as $\mathcal{O}(T)$ in Frobenius norm, causing progressive interference between stored associations. We introduce…

Machine Learning · Computer Science 2026-05-13 Vishal Pandey , Gopal Singh

The Transformer architecture, a cornerstone of modern Large Language Models (LLMs), has achieved extraordinary success in sequence modeling, primarily due to its attention mechanism. However, despite its power, the standard attention…

Machine Learning · Computer Science 2026-01-08 Zichuan Fu , Wentao Song , Guojing Li , Yejing Wang , Xian Wu , Yimin Deng , Hanyu Yan , Yefeng Zheng , Xiangyu Zhao

Transformer-based scientific foundation models are increasingly deployed in high-stakes settings, but current architectures give deterministic outputs and provide limited support for calibrated predictive uncertainty. We propose Stochastic…

Machine Learning · Computer Science 2026-05-12 Akash Yadav , Taiwo A. Adebiyi , Ruda Zhang

The quadratic computational complexity of softmax transformers has become a bottleneck in long-context scenarios. In contrast, linear attention model families provide a promising direction towards a more efficient sequential model. These…

Computation and Language · Computer Science 2026-02-04 Difan Deng , Andreas Bentzen Winje , Lukas Fehring , Marius Lindauer

The attention module, which is a crucial component in Transformer, cannot scale efficiently to long sequences due to its quadratic complexity. Many works focus on approximating the dot-then-exponentiate softmax function in the original…

Machine Learning · Computer Science 2021-11-04 Shengjie Luo , Shanda Li , Tianle Cai , Di He , Dinglan Peng , Shuxin Zheng , Guolin Ke , Liwei Wang , Tie-Yan Liu

The self-attention mechanism is the key to the success of transformers in recent Large Language Models (LLMs). However, the quadratic computational cost $O(n^2)$ in the input sequence length $n$ is a notorious obstacle for further…

Machine Learning · Computer Science 2024-10-17 Yingyu Liang , Heshan Liu , Zhenmei Shi , Zhao Song , Zhuoyan Xu , Junze Yin

Softmax attention defines an interaction through $d_h$ head dimensions, but not all dimensions carry equal weight once real text passes through. We decompose the attention logit field into a learned component and a generated component and…

Computation and Language · Computer Science 2026-04-09 Wonsuk Lee

The Hierarchical Kernel Transformer (HKT) is a multi-scale attention mechanism that processes sequences at L resolution levels via trainable causal downsampling, combining level-specific score matrices through learned convex weights. The…

Machine Learning · Computer Science 2026-04-13 Giansalvo Cirrincione

Neural networks in the lazy training regime converge to kernel machines. Can neural networks in the rich feature learning regime learn a kernel machine with a data-dependent kernel? We demonstrate that this can indeed happen due to a…

Machine Learning · Statistics 2022-02-07 Alexander Atanasov , Blake Bordelon , Cengiz Pehlevan

Linear attention methods offer a compelling alternative to softmax attention due to their efficiency in recurrent decoding. Recent research has focused on enhancing standard linear attention by incorporating gating while retaining its…

Machine Learning · Computer Science 2025-04-08 Yingcong Li , Davoud Ataee Tarzanagh , Ankit Singh Rawat , Maryam Fazel , Samet Oymak

Transformer models have been widely adopted in various domains over the last years, and especially large language models have advanced the field of AI significantly. Due to their size, the capability of these networks has increased…

Machine Learning · Computer Science 2023-11-10 Yelysei Bondarenko , Markus Nagel , Tijmen Blankevoort

Transformers often display an attention sink: probability mass concentrates on a fixed, content-agnostic position. Are sinks a byproduct of the optimization/training regime? Or are they sometimes functionally necessary in softmax…

Machine Learning · Computer Science 2026-04-20 Yuval Ran-Milo

In the classical transformer attention scheme, we are given three $n \times d$ size matrices $Q, K, V$ (the query, key, and value tokens), and the goal is to compute a new $n \times d$ size matrix $D^{-1} \exp(QK^\top) V$ where $D =…

Data Structures and Algorithms · Computer Science 2023-10-09 Josh Alman , Zhao Song