Metraipolis › archive › dc_35a22a016bb3

Correcting a learned physical invariant improves world-model rollouts — arXiv:2608.23526

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.23526
License (as archived)
CC0-1.0
Archived
2026-10-18T07:42:00.428Z
SHA-256 of the archived text
c0658b0ee557cd2f295fb4dde4ac6935e3f75629d8f63f32d4d5c59b2a76ef0f
Stable URL of this page
https://metraipolis.com/documento/dc_35a22a016bb3

How to cite

Correcting a learned physical invariant improves world-model rollouts — arXiv:2608.23526 — arXiv (en), https://arxiv.org/abs/2608.23526 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_35a22a016bb3 (Metraipolis archive dc_35a22a016bb3, sha256 c0658b0ee557cd2f295fb4dde4ac6935e3f75629d8f63f32d4d5c59b2a76ef0f)
BibTeX
@misc{dc35a22a016bb3,
  title = {Correcting a learned physical invariant improves world-model rollouts — arXiv:2608.23526},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.23526},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_35a22a016bb3; sha256 c0658b0ee557cd2f295fb4dde4ac6935e3f75629d8f63f32d4d5c59b2a76ef0f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.23526. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

World models can predict video without learning dynamics that they reliably preserve. We test whether a frozen DreamerV3 trained only on pendulum video learns a scalar that its own latent transition treats as approximately conserved. A label-free search recovers the same energy-like invariant across independently trained conservative models, while the same procedure finds no comparable invariant in matched damped models. During autonomous rollouts, this quantity drifts. Projecting the latent state back toward its initial level set reduces rollout error in all three conservative models, whereas matched random constraints usually increase it. These results distinguish a dynamically meaningful invariant from a merely decodable correlate and reveal a concrete failure mode: a world model can learn a physical constraint from pixels yet violate that constraint when it imagines forward.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-25 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.23526 — 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 2026-10-18T07:37:52.784Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).