arXiv — NLP / Computation & Language · · 3 min read

Top-Theta Attention: Sparsifying Transformers by Compensated Thresholding

Mirrored from arXiv — NLP / Computation & Language for archival readability. Support the source by reading on the original site.

Computer Science > Computation and Language

arXiv:2502.08363 (cs)
[Submitted on 12 Feb 2025 (v1), last revised 16 Jun 2026 (this version, v3)]

Title:Top-Theta Attention: Sparsifying Transformers by Compensated Thresholding

View a PDF of the paper titled Top-Theta Attention: Sparsifying Transformers by Compensated Thresholding, by Konstantin Berestizshevsky and 2 other authors
View PDF HTML (experimental)
Abstract:We present Top-Theta (Top-$\theta$) Attention, a training-free method for sparsifying transformer attention during inference. Our key insight is that static, per-head thresholds can be calibrated to retain the desired constant number of significant elements per attention row. This approach enables content-based sparsity without retraining, and it remains robust across data domains. We further introduce compensation techniques to preserve accuracy under aggressive sparsification, establishing attention thresholding as a practical and principled alternative to top-k attention. We provide extensive evaluation on natural language processing tasks, showing that Top-$\theta$ achieves 3-10x reduction in V-cache usage and up to 10x fewer attention elements during inference while degrading no more than 1% in accuracy.
Comments: Extended version of a paper accepted at ICANN 2026
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
MSC classes: 68T01
ACM classes: I.2
Cite as: arXiv:2502.08363 [cs.CL]
  (or arXiv:2502.08363v3 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2502.08363
arXiv-issued DOI via DataCite

Submission history

From: Konstantin Berestizshevsky [view email]
[v1] Wed, 12 Feb 2025 12:50:15 UTC (9,569 KB)
[v2] Fri, 22 Aug 2025 09:24:39 UTC (4,712 KB)
[v3] Tue, 16 Jun 2026 14:51:28 UTC (4,707 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled Top-Theta Attention: Sparsifying Transformers by Compensated Thresholding, by Konstantin Berestizshevsky and 2 other authors
  • View PDF
  • HTML (experimental)
  • TeX Source

Current browse context:

cs.CL
< prev   |   next >
Change to browse by:

References & Citations

Loading...

BibTeX formatted citation

loading...
Data provided by:

Bookmark

BibSonomy Reddit
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?)
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.

Discussion (0)

Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.

Sign in →

No comments yet. Sign in and be the first to say something.

More from arXiv — NLP / Computation & Language