Metraipolis › archive › dc_4479e61cfd24
Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Models — arXiv:2609.00355
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.00355
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T06:31:48.602Z
- SHA-256 of the archived text
959884335ab602eb4844dc5e7fa7f468c77169c2eb52c9a00250a931fe346854- Stable URL of this page
https://metraipolis.com/documento/dc_4479e61cfd24
How to cite
Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Models — arXiv:2609.00355 — arXiv (en), https://arxiv.org/abs/2609.00355 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4479e61cfd24 (Metraipolis archive dc_4479e61cfd24, sha256 959884335ab602eb4844dc5e7fa7f468c77169c2eb52c9a00250a931fe346854)
BibTeX
@misc{dc4479e61cfd24,
title = {Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Models — arXiv:2609.00355},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.00355},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4479e61cfd24; sha256 959884335ab602eb4844dc5e7fa7f468c77169c2eb52c9a00250a931fe346854},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.00355. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Speculative decoding accelerates generation without changing its output, but on vision-language models (VLMs) a self-reinforcing cycle holds it back. Because an autoregressive drafter pays a sequential pass for each drafted token, it must stay small and can ill afford to attend to the image at each pass. Prior work therefore compresses or hides the image, leaving the drafter weakest on the text the image determines. We present GLANCE, a one-pass block drafter that breaks this cycle on an unmodified VLM target. Its block-diffusion head drafts a whole block in one forward pass over the target's already fused vision-language states, reading the multimodal context once, however deep the draft. The target verifies a wide candidate tree in one pass and commits exactly its greedy output. In one production engine at a fixed round budget, GLANCE decodes up to 3.05 times faster than autoregressive decoding and outpaces the production EAGLE3-VL head on average and by about 11% on grounded tasks. An entropy law explains when drafting pays, predicting the longest accepted blocks on grounded tasks, where the target's next-token entropy is lowest. Our code is available at this https URL . Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-02 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2609.00355 — 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 2026-10-17T06:15:12.591Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).