Metraipolis › archive › dc_e4e8b5667674
LumoTree: Path-Parallel Speculative Verification for Hybrid Language Models — arXiv:2609.23900
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.23900
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-09T07:07:26.146Z
- SHA-256 of the archived text
7c7bf18fab8bdad76eb3985e11ee5e8175ad171c24d87dd7bf875771815f1794- Stable URL of this page
https://metraipolis.com/documento/dc_e4e8b5667674
How to cite
LumoTree: Path-Parallel Speculative Verification for Hybrid Language Models — arXiv:2609.23900 — arXiv (en), https://arxiv.org/abs/2609.23900 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e4e8b5667674 (Metraipolis archive dc_e4e8b5667674, sha256 7c7bf18fab8bdad76eb3985e11ee5e8175ad171c24d87dd7bf875771815f1794)
BibTeX
@misc{dce4e8b5667674,
title = {LumoTree: Path-Parallel Speculative Verification for Hybrid Language Models — arXiv:2609.23900},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.23900},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e4e8b5667674; sha256 7c7bf18fab8bdad76eb3985e11ee5e8175ad171c24d87dd7bf875771815f1794},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.23900. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Tree speculative decoding for recurrent-hybrid language models requires each accepted path to maintain consistent recurrent state, convolution history, and attention caches. We present LumoTree, a GPU serving design that coordinates verification and commitment through a shared tree descriptor. The verifier processes independent paths in parallel and reuses recurrent state tiles across local updates. Boundary states connect dependent paths. Accepted-path replay publishes the selected continuation into native running state, alongside coordinated convolution gathering and attention-cache remapping. Fused selection, device-resident acceptance, graph replay, and tree-aware attention integrate this organization into the serving cycle. The design separates temporary branch computation from persistent request state while retaining native prefix-cache interfaces. We evaluate numerical behavior, serving performance, and coding-agent outcomes on NVIDIA DGX Spark. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-22 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.23900 — 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 2026-10-09T06:58:09.347Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).