Structured sparse attention for accelerating long-context diffusion language models

Introduction Diffusion Large Language Models (DLLMs) have emerged as a compelling parallel decoding paradigm. However, the inherent full bidirectional attention at every denoising step incurs quadratic complexity, imposing severe latency bottlenecks in long-context scenarios. Methods We investigate attention dynamics in DLLMs and propose Structured Sparse Attention (SSA), which uses a fast structure-wise algorithm to select and reuse geometric patterns for inference acceleration. SSA combines cached structured patterns with protection of the active decoding region to balance inference efficiency and output quality. Results Our investigation reveals a decrease in average adjacent-step attention changes later in denoising, together with vertical and diagonal geometric patterns that motivate sparse-pattern reuse; the trend varies across inputs, layers, and heads. Extensive experiments demonstrate that SSA delivers significant speedups while maintaining competitive quality against previous state-of-the-art methods on the LongBench and RULER benchmarks. At a 128K context length, SSA achieves an end-to-end inference speedup of approximately 2 × . Discussion These findings support structured-pattern reuse combined with decoding-region protection as an efficient, training-free approach to accelerating long-context DLLM inference.

Authors

Institutions

Publication Details

Journal
Frontiers in Artificial Intelligence
Published
2026-11-18
DOI
https://doi.org/10.3389/frai.2026.1993541
Primary Topic
Topic Modeling
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Structured sparse attention for accelerating long-context diffusion language models

Haoyun Jiang, Xiangtao Li, Feng Hong, Yicheng Fu et al.
Frontiers in Artificial Intelligence
Topic Modeling
article

Structured sparse attention for accelerating long-context diffusion language models

Haoyun Jiang, Xiangtao Li, Feng Hong, Yicheng Fu, Pengyi Chen, Xinmu Ge, Bo Han, Jiangchao Yao, Ziyi Tang, Xu Chen, Xiaofeng Cao
article en

Abstract

Introduction Diffusion Large Language Models (DLLMs) have emerged as a compelling parallel decoding paradigm. However, the inherent full bidirectional attention at every denoising step incurs quadratic complexity, imposing severe latency bottlenecks in long-context scenarios. Methods We investigate attention dynamics in DLLMs and propose Structured Sparse Attention (SSA), which uses a fast structure-wise algorithm to select and reuse geometric patterns for inference acceleration. SSA combines cached structured patterns with protection of the active decoding region to balance inference efficiency and output quality. Results Our investigation reveals a decrease in average adjacent-step attention changes later in denoising, together with vertical and diagonal geometric patterns that motivate sparse-pattern reuse; the trend varies across inputs, layers, and heads. Extensive experiments demonstrate that SSA delivers significant speedups while maintaining competitive quality against previous state-of-the-art methods on the LongBench and RULER benchmarks. At a 128K context length, SSA achieves an end-to-end inference speedup of approximately 2 × . Discussion These findings support structured-pattern reuse combined with decoding-region protection as an efficient, training-free approach to accelerating long-context DLLM inference.

Frontiers in Artificial IntelligenceVol. 9
Tongji University (CN), Hong Kong Baptist University (HK), Shanghai Jiao Tong University (CN), Northeast Normal University (CN), Shanghai Innovation Institute (CN), Alibaba Group (China) (CN)
Topic Modeling
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon