Metraipolis › archive › dc_67b4313fef9e
world2vec: Mechanistic world models through latent dynamics learning — arXiv:2610.40637
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.40637
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-01T02:26:44.618Z
- SHA-256 of the archived text
6c3d225efc42b8e5e509447a68f9d869a2e08438598cd75ff55dc2a7359f829d- Stable URL of this page
https://metraipolis.com/documento/dc_67b4313fef9e
How to cite
world2vec: Mechanistic world models through latent dynamics learning — arXiv:2610.40637 — arXiv (en), https://arxiv.org/abs/2610.40637 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_67b4313fef9e (Metraipolis archive dc_67b4313fef9e, sha256 6c3d225efc42b8e5e509447a68f9d869a2e08438598cd75ff55dc2a7359f829d)
BibTeX
@misc{dc67b4313fef9e,
title = {world2vec: Mechanistic world models through latent dynamics learning — arXiv:2610.40637},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.40637},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_67b4313fef9e; sha256 6c3d225efc42b8e5e509447a68f9d869a2e08438598cd75ff55dc2a7359f829d},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.40637. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
World models can achieve strong control performance without faithfully capturing the dynamics of their environment, limiting their potential for mechanistic interpretability and scientific discovery. Here, we introduce world2vec to learn mechanistic world models without reward information directly from observations and actions. World2vec learns latent dynamics that faithfully reproduce the true system over long prediction horizons, with at least one order of magnitude fewer parameters than DreamerV3, TD-MPC2 and LeWM. This enables state of the art performance in controlling continuous environments from pixels and accurate system identification across a benchmark of 71 attractors, demonstrating how world models are amenable to scientific discovery. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.40637 — 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 2026-11-01T02:12:44.223Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).