aiqbench analysis
This paper introduces MOSH-WM, a novel object-centric world model that improves video forecasting by explicitly grounding its state variables in image support masks rather than unconstrained features. By integrating soft-Hamiltonian dynamics with causal visual contexts, the approach significantly reduces prediction error and slows error accumulation over extended horizons on standard benchmarks like OBJ3D and CLEVRER. The work advances the field of generative AI by providing a more robust framework for simulating physical object interactions in future video sequences.
MOSH-WM: Mask-Grounded Soft-Hamiltonian Dynamics for Object-Centric World Models
arXiv:2608.22750v1 Announce Type: new Abstract: Object-centric world models forecast future videos by evolving a set of entity slots, but the variables receiving dynamics supervision are often unconstrained visual features. We introduce \method{}, a mask-grounded soft-Hamiltonian world model that makes its position-like state explicitly depend on slot-owned image support. A frozen video-slot encoder produces slots and masks; spatial moments of mask-owned support form a canonical state $Q$, temporal differences form $P$, and a learned energy supplies a soft directional bias to a bounded learned increment. Decoder-relevant appearance and identity are stored separately in a causal visual context. A gated composer and bounded residual then combine this context with the propagated phase state to reconstruct decoder-compatible slots. On OBJ3D, given six observed frames and evaluated over the following 30 frames, \method{} reduces LPIPS by 25.0\% and spatial MSE by 33.7\% relative to the strongest object-centric baseline. On CLEVRER, given six observed frames and evaluated over the following ten frames, the corresponding reductions are 14.5\% and 18.7\%. Horizon-resolved visual and object-state measurements show that the complete model accumulates error more slowly throughout the 30-frame closed-loop rollout. Project page:https://github.com/moshwm-anon/-moshwm-anon.github.io.
Abstract:Object-centric world models forecast future videos by evolving a set of entity slots, but the variables receiving dynamics supervision are often unconstrained visual features. We introduce \method{}, a mask-grounded soft-Hamiltonian world model that makes its position-like state explicitly depend on slot-owned image support. A frozen video-slot encoder produces slots and masks; spatial moments of mask-owned support form a canonical state $Q$, temporal differences form $P$, and a learned energy supplies a soft directional bias to a bounded learned increment. Decoder-relevant appearance and identity are stored separately in a causal visual context. A gated composer and bounded residual then combine this context with the propagated phase state to reconstruct decoder-compatible slots. On OBJ3D, given six observed frames and evaluated over the following 30 frames, \method{} reduces LPIPS by 25.0\% and spatial MSE by 33.7\% relative to the strongest object-centric baseline. On CLEVRER, given six observed frames and evaluated over the following ten frames, the corresponding reductions are 14.5\% and 18.7\%. Horizon-resolved visual and object-state measurements show that the complete model accumulates error more slowly throughout the 30-frame closed-loop rollout. Project page:this https URL.
Machine Learning (cs.LG)
arXiv:2608.22750 [cs.LG]
(or arXiv:2608.22750v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.22750 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 276 words
Related coverage
Beyond Instance Slots: Semantically Rich World Models for Physical Interaction Planning
arXiv · Aug 25, 2026
Soft Geometric Inductive Bias for Object Centric Dynamics
arXiv · Sep 30, 2026
AcrossVAM1.0: Particle World Modeling for Text-Assisted Robot Video Prediction
arXiv · Sep 1, 2026
[AINews] not much happened today
Latent Space · Sep 10, 2026