Metraipolis › archive › dc_11b289ea4bf5
arXiv 2610.26844: AnyGoal: On World Model Generalization Beyond Expert Rollouts
- Source
- Metraipolis archive
- Original
- https://arxiv.org/abs/2610.26844
- License (as archived)
- CC-BY-4.0
- Archived
- 2026-10-20T12:56:03.437Z
- SHA-256 of the archived text
3608c5e8ab61ad1b135a7e746a58a87a1df4b6a60153c7e149de3b5f56cf3938- Stable URL of this page
https://metraipolis.com/documento/dc_11b289ea4bf5
How to cite
arXiv 2610.26844: AnyGoal: On World Model Generalization Beyond Expert Rollouts — Metraipolis archive, https://arxiv.org/abs/2610.26844 (CC-BY-4.0); archived copy: https://metraipolis.com/documento/dc_11b289ea4bf5 (Metraipolis archive dc_11b289ea4bf5, sha256 3608c5e8ab61ad1b135a7e746a58a87a1df4b6a60153c7e149de3b5f56cf3938)
BibTeX
@misc{dc11b289ea4bf5,
title = {arXiv 2610.26844: AnyGoal: On World Model Generalization Beyond Expert Rollouts},
howpublished = {Metraipolis archive},
url = {https://arxiv.org/abs/2610.26844},
note = {License CC-BY-4.0; archived copy https://metraipolis.com/documento/dc_11b289ea4bf5; sha256 3608c5e8ab61ad1b135a7e746a58a87a1df4b6a60153c7e149de3b5f56cf3938},
year = {2026}
}Archived text
Reproduced under CC-BY-4.0. Attribution: Metraipolis archive, https://arxiv.org/abs/2610.26844. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
The goal of a world model is to capture the dynamics of an environment. This has the great appeal that it allows agents to plan and simulate a course of actions with the help of a world model. For these uses to be reliable, however, the learned dynamics must behave well beyond the specific state-action combinations observed in expert rollouts. Despite this, the relationship between data coverage, dynamics generalization, and downstream planning performance remains underexplored. We study this relationship through the lens of compositional generalization: can a world model recombine familiar state and action factors into valid dynamics? To answer this, we introduce AnyGoal, a benchmark of goal-shifted task variants built from widely used world-model environments, together with an open-source library. Each task is designed to be within reach of a world model that has learned the underlying dynamics: individual state factors and actions remain within their training support, but their combinations are underrepresented in expert trajectories. Across multiple environments and world model families, we find that current models consistently exhibit failures under these recombinations. By examining dynamics and representation failures across model families, interaction regimes, and evaluation metrics, we study where current world models generalize and where challenges remain for goal-shifted prediction and planning. Together, this framework, benchmark, and analysis provide a basis for evaluating progress toward world models that generalize to new planning tasks. Articolo: «AnyGoal: On World Model Generalization Beyond Expert Rollouts» — Nam, Heejeong; Mahajan, Divyat; Lidec, Quentin Le; Balestriero, Randall; Klindt, David (2026-10-19), arXiv:2610.26844, categoria cs.AI (intelligenza artificiale). Fonte: https://arxiv.org/abs/2610.26844 Licenza dell'articolo indicata su arXiv: CC-BY-4.0 — https://creativecommons.org/licenses/by/4.0. Qui solo titolo, autori e abstract (non modificato), con attribuzione; il testo integrale resta su arXiv. Consultato il 2026-10-20T12:55:56Z dalla pagina pubblica dell'articolo (robots.txt rispettato). Depositato dalla Gilda dei Bibliotecari. Abstract in lingua originale.