[2605.18753] DashAttention: Differentiable and Adaptive Sparse Hierarchical Attention

-->

Computer Science > Computation and Language

arXiv:2605.18753 (cs)

[Submitted on 18 May 2026]

Title:DashAttention: Differentiable and Adaptive Sparse Hierarchical Attention

Authors:Yuxiang Huang, Nuno M. T. Gonçalves, Federico Alvetreti, Lei Li, Xu Han, Edoardo M. Ponti, André F. T. Martins, Marcos V. Treviso<br>View a PDF of the paper titled DashAttention: Differentiable and Adaptive Sparse Hierarchical Attention, by Yuxiang Huang and 7 other authors

View PDF

Abstract: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 top-k operation assumes the number of relevant tokens for any query is fixed and it precludes the gradient flow between the sparse and dense stages. In this work, we propose DashAttention (Differentiable and Adaptive Sparse Hierarchical Attention), which leverages the adaptively sparse $\alpha$-entmax transformation to select a variable number of blocks according to the current query in the first stage. This in turn provides a prior for the second-stage softmax attention, keeping the entire hierarchy fully differentiable. Contrary to other hierarchical attention methods, we show that DashAttention is non-dispersive, translating to better long-context modeling ability. Experiments with large language models (LLMs) show that DashAttention achieves comparable accuracy as full attention with 75% sparsity and a better Pareto frontier than NSA and InfLLMv2, especially in high-sparsity regimes. We also provide an efficient, GPU-aware implementation of DashAttention in Triton, which achieves a speedup of up to over FlashAttention-3 at inference time. Overall, DashAttention offers a cost-effective strategy to model long contexts.

Comments:<br>Preprint

Subjects:

Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)

Cite as:<br>arXiv:2605.18753 [cs.CL]

(or<br>arXiv:2605.18753v1 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2605.18753

Focus to learn more

arXiv-issued DOI via DataCite (pending registration)

Submission history<br>From: Yuxiang Huang [view email]<br>[v1]<br>Mon, 18 May 2026 17:59:52 UTC (936 KB)

Full-text links:<br>Access Paper:

View a PDF of the paper titled DashAttention: Differentiable and Adaptive Sparse Hierarchical Attention, by Yuxiang Huang and 7 other authors<br>View PDF<br>TeX Source

view license

Current browse context:

cs.CL

next >

new<br>recent<br>| 2026-05

Change to browse by:

cs<br>cs.AI<br>cs.LG

References & Citations

NASA ADS<br>Google Scholar

Semantic Scholar

export BibTeX citation<br>Loading...

BibTeX formatted citation

&times;

loading...

Data provided by:

Bookmark

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

Which authors of this paper are endorsers? |<br>Disable MathJax (What is MathJax?)