Mixture-of-Top-k Attention: Efficient Attention via Scalable Fast Weights
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866917476257759232 |
|---|---|
| author | Wen, Qishuai Huang, Zhiyuan Meng, Xianghan He, Wei Li, Chun-Guang |
| author_facet | Wen, Qishuai Huang, Zhiyuan Meng, Xianghan He, Wei Li, Chun-Guang |
| contents | The vanilla self-attention mechanism in Transformers can be viewed as a two-layer fast-weight MLP, whose weights are dynamically induced by inputs and whose hidden dimension is equal to the sequence length $N$. As the context extends, the expressive capacity of such an $N$-width MLP increases, but it becomes unscalable for extremely long sequences. Recently, this fast-weight perspective has motivated the Mixture-of-Experts (MoE) attention mechanism, which partitions the sequence into rigid blocks, treats them as fast-weight experts, and sparsely routes the tokens to them. In this paper, we elevate this perspective to a unifying framework for efficient attention mechanisms, interpreting them as making fast weights scalable through either routing or compression, and organizing them into a five-dimensional taxonomy. Then, we propose Mixture-of-Top-$k$ Attention (MiTA), which employs a small set of landmark queries to gather top-$k$ attended key-value pairs as query-aware and deformable routed experts, while compressing the $N$-width MLP into a narrower shared expert. Consequently, our MiTA improves the flexibility of prior MoE attention from rigid to deformable fast-weight experts, as well as the scalability of prior top-$k$ attention from query-specific set to reusable top-$k$ set. We conduct extensive experiments on vision tasks showing the superior effectiveness and efficiency of our MiTA, and also uncovering intriguing properties such as an emergent token-pruning effect and easy generalization from standard attention. Code is available at https://github.com/QishuaiWen/MiTA. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_01219 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Mixture-of-Top-k Attention: Efficient Attention via Scalable Fast Weights Wen, Qishuai Huang, Zhiyuan Meng, Xianghan He, Wei Li, Chun-Guang Machine Learning Computer Vision and Pattern Recognition The vanilla self-attention mechanism in Transformers can be viewed as a two-layer fast-weight MLP, whose weights are dynamically induced by inputs and whose hidden dimension is equal to the sequence length $N$. As the context extends, the expressive capacity of such an $N$-width MLP increases, but it becomes unscalable for extremely long sequences. Recently, this fast-weight perspective has motivated the Mixture-of-Experts (MoE) attention mechanism, which partitions the sequence into rigid blocks, treats them as fast-weight experts, and sparsely routes the tokens to them. In this paper, we elevate this perspective to a unifying framework for efficient attention mechanisms, interpreting them as making fast weights scalable through either routing or compression, and organizing them into a five-dimensional taxonomy. Then, we propose Mixture-of-Top-$k$ Attention (MiTA), which employs a small set of landmark queries to gather top-$k$ attended key-value pairs as query-aware and deformable routed experts, while compressing the $N$-width MLP into a narrower shared expert. Consequently, our MiTA improves the flexibility of prior MoE attention from rigid to deformable fast-weight experts, as well as the scalability of prior top-$k$ attention from query-specific set to reusable top-$k$ set. We conduct extensive experiments on vision tasks showing the superior effectiveness and efficiency of our MiTA, and also uncovering intriguing properties such as an emergent token-pruning effect and easy generalization from standard attention. Code is available at https://github.com/QishuaiWen/MiTA. |
| title | Mixture-of-Top-k Attention: Efficient Attention via Scalable Fast Weights |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.01219 |