$π$-Attention: Periodic Sparse Transformers for Efficient Long-Context Modeling

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Dong, Yu, Yanxuan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918413837795328
author Liu, Dong
Yu, Yanxuan
author_facet Liu, Dong
Yu, Yanxuan
contents Transformers have revolutionized natural language processing, but their quadratic complexity with respect to sequence length remains a fundamental bottleneck for long-range modeling. While sparse attention mechanisms like RingAttention reduce computational costs by restricting attention to local neighborhoods, they suffer from limited receptive fields and lack of adaptability. We present \PiAttention, a periodic sparse Transformer that factorizes attention into ring-local neighborhoods, deterministic $π$-stride skips, and an adaptive fusion gate. The periodic structure provides predictable coverage of distant tokens, while the sparse footprint keeps the per-layer complexity linear in context length. We prove that \PiAttention achieves $\mathcal{O}(kL + π\log L)$ receptive field growth compared to $\mathcal{O}(kL)$ for RingAttention, where $k$ is the local window size, $π$ is the skip period, and $L$ is the sequence length. Extensive experiments on language modeling, retrieval, and vision-language tasks demonstrate that \PiAttention matches or surpasses dense attention quality with 8.3\% lower perplexity than RingAttention while using 50\% fewer GPUs for the same context length. Our detailed ablations and visualizations reveal the importance of periodic skips, adaptive fusion, and head-level sparsity coordination for efficient long-context modeling.
format Preprint
id arxiv_https___arxiv_org_abs_2511_10696
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle $π$-Attention: Periodic Sparse Transformers for Efficient Long-Context Modeling
Liu, Dong
Yu, Yanxuan
Computation and Language
Artificial Intelligence
Transformers have revolutionized natural language processing, but their quadratic complexity with respect to sequence length remains a fundamental bottleneck for long-range modeling. While sparse attention mechanisms like RingAttention reduce computational costs by restricting attention to local neighborhoods, they suffer from limited receptive fields and lack of adaptability. We present \PiAttention, a periodic sparse Transformer that factorizes attention into ring-local neighborhoods, deterministic $π$-stride skips, and an adaptive fusion gate. The periodic structure provides predictable coverage of distant tokens, while the sparse footprint keeps the per-layer complexity linear in context length. We prove that \PiAttention achieves $\mathcal{O}(kL + π\log L)$ receptive field growth compared to $\mathcal{O}(kL)$ for RingAttention, where $k$ is the local window size, $π$ is the skip period, and $L$ is the sequence length. Extensive experiments on language modeling, retrieval, and vision-language tasks demonstrate that \PiAttention matches or surpasses dense attention quality with 8.3\% lower perplexity than RingAttention while using 50\% fewer GPUs for the same context length. Our detailed ablations and visualizations reveal the importance of periodic skips, adaptive fusion, and head-level sparsity coordination for efficient long-context modeling.
title $π$-Attention: Periodic Sparse Transformers for Efficient Long-Context Modeling
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2511.10696