Metraipolis › archive › dc_0bbb608e64ef

HyScale: Scaling Long-Context Training for Learned Sparse Attention — arXiv:2610.32057

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.32057
License (as archived): CC0-1.0
Archived: 2026-10-26T19:47:39.870Z
SHA-256 of the archived text: f40ac2eaa62ac16a8119e061cab1dba718d9a13edc08dd1bc245bcde74c8e5ed
Stable URL of this page: https://metraipolis.com/documento/dc_0bbb608e64ef

How to cite

HyScale: Scaling Long-Context Training for Learned Sparse Attention — arXiv:2610.32057 — arXiv (en), https://arxiv.org/abs/2610.32057 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0bbb608e64ef (Metraipolis archive dc_0bbb608e64ef, sha256 f40ac2eaa62ac16a8119e061cab1dba718d9a13edc08dd1bc245bcde74c8e5ed)

BibTeX

@misc{dc0bbb608e64ef,
  title = {HyScale: Scaling Long-Context Training for Learned Sparse Attention — arXiv:2610.32057},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.32057},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0bbb608e64ef; sha256 f40ac2eaa62ac16a8119e061cab1dba718d9a13edc08dd1bc245bcde74c8e5ed},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.32057. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We present HyScale, a long-context training system for learned sparse attention. Learned sparse attention is a promising architecture for million-token contexts, widely adopted in production since DeepSeek-V3.2. Selection, however, grows into the training bottleneck. For each query, it scores every previous token whereas sparse attention reads only $k$; worse, current systems materialize a full score table and repeat selection during recomputation. Our key insight is that the output of selection is not an activation, but rather a set of indices. Building on it, HyScale introduces HyTopK and selection replay. HyTopK finds each query's exact top-$k$ on the fly without a score table, while selection replay reuses forward-pass selection in recomputation. Together with holistic system optimizations tackling end-to-end training challenges, HyScale speeds up GLM-5.1 training on 1M-token sequences by 2.07$\times$ over Megatron-LM on 512 H20 GPUs. It also speeds up the CSA layers of DeepSeek-V4-Flash and the QSA layers of Qwen3.8-Flash-Next by 2.01$\times$ and 1.57$\times$ at 1M tokens on 32 H800 GPUs.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Distributed, Parallel, and Cluster Computing (cs.DC).
Pagina del lavoro: https://arxiv.org/abs/2610.32057 — 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 2026-10-26T19:45:33.280Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).