Metraipolis › archive › dc_6ca7bbba0599
HyFlash: A Prefix-Memory Prior for On-Device Speculative Decoding — arXiv:2611.07263
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.07263
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-06T14:43:04.837Z
- SHA-256 of the archived text
408c053fdad0a02811cf72818a9bbd0ff492bad8ccb4a3884b7fd873e2f84226- Stable URL of this page
https://metraipolis.com/documento/dc_6ca7bbba0599
How to cite
HyFlash: A Prefix-Memory Prior for On-Device Speculative Decoding — arXiv:2611.07263 — arXiv (en), https://arxiv.org/abs/2611.07263 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6ca7bbba0599 (Metraipolis archive dc_6ca7bbba0599, sha256 408c053fdad0a02811cf72818a9bbd0ff492bad8ccb4a3884b7fd873e2f84226)
BibTeX
@misc{dc6ca7bbba0599,
title = {HyFlash: A Prefix-Memory Prior for On-Device Speculative Decoding — arXiv:2611.07263},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.07263},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6ca7bbba0599; sha256 408c053fdad0a02811cf72818a9bbd0ff492bad8ccb4a3884b7fd873e2f84226},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.07263. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Parallel block drafting accelerates speculative decoding by predicting multiple future positions in a single backbone pass, but on-device deployment constrains both drafter capacity and context storage. Under tight memory budgets, drafters must model long-range dependencies and fine-grained draft-position interactions without the growing KV-cache overhead of full-context attention. We propose HyFlash, a compact parallel drafter that introduces a prefix-memory prior to separate long-range context modeling from fine-grained draft-position interactions with limited capacity. Local attention captures recent context and draft-position interactions, while recurrent memory compresses historical target features into fixed-size states. This inductive bias enables effective local and long-range modeling with bounded draft-side memory. Experiments show that HyFlash achieves higher acceptance than full-context attention with substantially less KV storage and with larger gains for shallow drafters and long contexts. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2611.07263 — 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 2026-11-06T14:18:57.584Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).