Metraipolis › archive › dc_d4801d8d7697

Beyond a single latent space: a dual-latent world model for long-horizon planning — arXiv:2609.37644

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.37644
License (as archived)
CC0-1.0
Archived
2026-10-07T16:24:00.080Z
SHA-256 of the archived text
1964664f2d9c6ba41173106931bc3d9360b75f228ea49e85305205b71cd0d0cc
Stable URL of this page
https://metraipolis.com/documento/dc_d4801d8d7697

How to cite

Beyond a single latent space: a dual-latent world model for long-horizon planning — arXiv:2609.37644 — arXiv (en), https://arxiv.org/abs/2609.37644 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d4801d8d7697 (Metraipolis archive dc_d4801d8d7697, sha256 1964664f2d9c6ba41173106931bc3d9360b75f228ea49e85305205b71cd0d0cc)
BibTeX
@misc{dcd4801d8d7697,
  title = {Beyond a single latent space: a dual-latent world model for long-horizon planning — arXiv:2609.37644},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.37644},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d4801d8d7697; sha256 1964664f2d9c6ba41173106931bc3d9360b75f228ea49e85305205b71cd0d0cc},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.37644. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Latent world models often struggle with long-horizon planning despite accurate short-term predictions. Recursive rollouts accumulate errors, while distance concentration in high-dimensional latent spaces can weaken goal discrimination. We introduce the Dual-Latent World Model (Dual-WM), which separates local execution and long-range planning through distinct state representations and dynamics models. The low-level model predicts action-conditioned transitions, while the high-level model uses learned macro-actions to plan over longer temporal spans. We also propose Long-Horizon Representation Learning with Weighted Rollout (LoRe), which supervises self-generated predictions at both levels. An analysis of recursive error propagation motivates exponential horizon weights with separate decay rates for the two temporal scales. During planning, the high-level model generates latent subgoals that the low-level model refines into actions for precise execution. We evaluate from-scratch Dual-WM on five goal-conditioned visual control tasks against the task-wise strongest baselines without actor-guided proposals. At goal offsets of 50 and 100 environment steps, mean success increases from 75.9% to 84.4% and from 61.4% to 69.5%, respectively. At offset 100, Dual-WM outperforms these baselines on all five tasks and improves mean success over LeWM by 30.8 percentage points. Ablations and supporting analyses provide evidence of more informative representations for goal evaluation and greater consistency under recursive prediction. These results highlight the value of separating temporal roles and training across multiple horizons for reliable latent planning. Our core implementation is available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.37644 — 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 2026-10-07T15:54:53.640Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).