Metraipolis › archive › dc_85693cbe89d8

SketchSSM: Write to the Full State, Read from a Compact Sketch — arXiv:2609.33051

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.33051
License (as archived)
CC0-1.0
Archived
2026-10-07T22:12:22.468Z
SHA-256 of the archived text
e9ef71baebce1c16ce62052c74087b87bd6102edb704309cfe4aa9ee12a8cfa0
Stable URL of this page
https://metraipolis.com/documento/dc_85693cbe89d8

How to cite

SketchSSM: Write to the Full State, Read from a Compact Sketch — arXiv:2609.33051 — arXiv (en), https://arxiv.org/abs/2609.33051 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_85693cbe89d8 (Metraipolis archive dc_85693cbe89d8, sha256 e9ef71baebce1c16ce62052c74087b87bd6102edb704309cfe4aa9ee12a8cfa0)
BibTeX
@misc{dc85693cbe89d8,
  title = {SketchSSM: Write to the Full State, Read from a Compact Sketch — arXiv:2609.33051},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.33051},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_85693cbe89d8; sha256 e9ef71baebce1c16ce62052c74087b87bd6102edb704309cfe4aa9ee12a8cfa0},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33051. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Hybrid-attention models replace most softmax attention layers with linear attention, reducing KV-cache growth and enabling larger decode batches where recurrent-state access becomes a major bottleneck. ReplaySSM amortizes state updates by buffering keys and values, but each new query still requires a full-state read even though the state remains unchanged between state updates. We observe that low-rank state-weighted query approximation accurately preserves state-read outputs. Although future queries are unknown, the basis vectors used to approximate them can be fixed offline. Based on this observation, we introduce SketchSSM, which preserves full-state updates while approximating reads. At each state update, SketchSSM reads the full state once to precompute outputs for these basis vectors, storing them in a compact sketch. Each subsequent decode step combines the sketch vectors with query-dependent coefficients to reconstruct the output without a full-state read. Across four Mamba-2-, GDN-, and KDA-based models, SketchSSM at a mean sketch rank of 8 reduces state-access traffic by approximately 10$\times$ while matching the average accuracy of the FP32 full-state baseline across four decode benchmarks, and preserves recall on four RULER retrieval tasks. At this rank on one NVIDIA B300, linear-attention kernel speedups over the Standard vLLM baseline reach 7.30$\times$, 5.02$\times$, and 5.24$\times$ for Mamba-2, GDN, and KDA, respectively, with up to 2.77$\times$ higher decode throughput on Nemotron 3 Super.

Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-29 in cs.DC).
Categorie arXiv: Machine Learning (cs.LG) ; Distributed, Parallel, and Cluster Computing (cs.DC).
Pagina del lavoro: https://arxiv.org/abs/2609.33051 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T21:53:13.764Z dalla pagina https://arxiv.org/catchup/cs.DC/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).