Metraipolis › archive › dc_19b2df32216e
Memorizon: Training World Models Beyond Their Context Window — arXiv:2610.00544
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.00544
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T19:11:06.450Z
- SHA-256 of the archived text
b9e1781897c90008acd17a6653e9381313256d6dc0653554e8a79e6e9a9bf32a- Stable URL of this page
https://metraipolis.com/documento/dc_19b2df32216e
How to cite
Memorizon: Training World Models Beyond Their Context Window — arXiv:2610.00544 — arXiv (en), https://arxiv.org/abs/2610.00544 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_19b2df32216e (Metraipolis archive dc_19b2df32216e, sha256 b9e1781897c90008acd17a6653e9381313256d6dc0653554e8a79e6e9a9bf32a)
BibTeX
@misc{dc19b2df32216e,
title = {Memorizon: Training World Models Beyond Their Context Window — arXiv:2610.00544},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.00544},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_19b2df32216e; sha256 b9e1781897c90008acd17a6653e9381313256d6dc0653554e8a79e6e9a9bf32a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.00544. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Streaming world models should render a place consistently across repeated visits. Directly supervising such revisits requires training samples that capture both visits, often spanning minutes. Yet dense attention over the full span incurs quadratic costs, making long-span supervision expensive. Memorizon breaks this coupling: long spans are needed for supervision, but not for attention, since the two visits can share a forward pass without including every intervening frame. A training sample covers a span of any length but is scored only on its last $k$ chunks. Instead of tokenizing the history before them, each scored chunk retrieves its own top-$K$ latents by camera co-visibility, and the union of these requests forms a shared bank. The bank is bounded by $kK$, so the sequence stays bounded however long the span; at the shortest span the recipe is exactly conventional training. Adding the bank raises the cost of a step once; beyond that, a longer span costs little, and going from 100 to 400 s adds 12% to the step time. Against a sliding-window baseline, retrieval raises revisit consistency on every split, and a span long enough to reach the first visit of each return adds a further 24% to 30%, at some cost in image quality; beyond that span, more length no longer helps. Filling the bank from another episode lowers revisit correlation by 83%, so the model uses what it retrieves. Project page: this https URL Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-10-02 in cs.CV). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2610.00544 — 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 2026-10-06T17:37:36.671Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).