Skip to content
AI News HubLIVE
Original source2 min read

ADIndex: Layer-Adaptive Sparse Attention Indexing via Group-Level Distillation

Summary

ADIndex is an adaptive, distillation-driven indexing method for token-level sparse attention that combines layer-wise calibration, group-level distillation, and low-latency inference to significantly reduce indexer latency in long-context inference, while maintaining near-lossless task performance on DeepSeek-V3.2 and GLM-5.1.

SourcearXiv Machine LearningAuthor: Jin-Fei Qi, Yiduo Li, Jiang-Xin Shi, Kan Liu, Tao Lan, Hanlin Tang, Yu-Feng Li
ADIndex: Layer-Adaptive Sparse Attention Indexing via Group-Level Distillation
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 7 Nov 2026]

Title:ADIndex: Layer-Adaptive Sparse Attention Indexing via Group-Level Distillation

View a PDF of the paper titled ADIndex: Layer-Adaptive Sparse Attention Indexing via Group-Level Distillation, by Jin-Fei Qi and 6 other authors

View PDF HTML (experimental)

Abstract:DeepSeek Sparse Attention (DSA) reduces the computational cost of attention through token-level selection. However, its indexer still scores the entire causal prefix for each query, making indexing a dominant source of latency in long-context inference. Notably, the indexer alone can account for more than 50% of total prefill GPU time at long context lengths. To address this bottleneck, we propose ADIndex, an Adaptive and Distillation-driven Indexing method for token-level sparse attention. ADIndex integrates three key components: (i) a layer-wise calibration strategy that adaptively accommodates cross-query retrieval overlap, motivated by our observation of pronounced inter-layer heterogeneity; (ii) a group-level distillation scheme that leverages this heterogeneity to efficiently train adaptive layer-wise group indexers, capturing salient retrieval signals from individual queries; and (iii) a low-latency inference strategy that reduces repeated full-prefix scans through token bank sharing within query groups while retaining query-specific refinement. Experiments on DeepSeek-V3.2 and GLM-5.1 demonstrate that ADIndex preserves near-lossless task performance on LongBench, LongBench v2, and RULER while delivering substantial efficiency gains over standard DSA, including prefill and decoding kernel speedups of up to 6.22x and 3.40x, respectively, at a context length of 256K across the layer-wise reuse configurations, as well as an end-to-end serving speedup of 1.54x at a context length of 128K.

Comments: 27 pages,13 figures

Subjects:

Machine Learning (cs.LG); Artificial Intelligence (cs.AI)

Cite as: arXiv:2611.12956 [cs.LG]

(or arXiv:2611.12956v1 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2611.12956

arXiv-issued DOI via DataCite

Submission history

From: Jinfei Qi [view email] [v1] Sat, 7 Nov 2026 06:35:09 UTC (1,084 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled ADIndex: Layer-Adaptive Sparse Attention Indexing via Group-Level Distillation, by Jin-Fei Qi and 6 other authors

View PDF

HTML (experimental)

TeX Source

view license

Additional Features

Audio Summary

Current browse context:

cs.LG

new | recent | 2026-11

Change to browse by:

cs cs.AI

References & Citations

SciX

Google Scholar

Semantic Scholar

Loading...

Data provided by:

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?)

IArxiv recommender toggle

IArxiv Recommender (What is IArxiv?)

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? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • The DSA indexer can account for over 50% of prefill GPU time at long context lengths.
  • ADIndex integrates layer-wise calibration, group-level distillation, and a low-latency inference strategy.
  • Achieves up to 6.22x prefill and 3.40x decoding kernel speedups at 256K context length.
  • Delivers 1.54x end-to-end serving speedup at 128K context length.

Highlights and analysis are generated automatically and may contain errors. Check the original source.