The Dual-Latent World Model is introduced, which separates local execution and long-range planning through distinct state representations and dynamics models and proposes Long-Horizon Representation Learning with Weighted Rollout (LoRe), which supervises self-generated predictions at both levels.
Abstract
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 https://github.com/DeLin1001/Dual-WM-Official.
H-JEPA is introduced, an end-to-end recipe for training a hierarchy of action-conditioned JEPAs in which each level predicts farther ahead in its own learned latent space in which each level's predictions become subgoals for the planner below it.
Wan-Cong Zhang, Basile Terver, Michael G. Rabbat et al.· 7 citations· ⚡2
Metro-WM is introduced, a hierarchical framework that issues sub-goals by retrieving genuine states from prior experience rather than generating ungrounded latent vectors that achieves superior long-horizon success rates and maintains robust performance under extremely sparse dataset conditions.
Royson Lee, Fady Rezk, Titouan Parcollet et al.· 0 citations
WALL-SS is introduced, a world model that generates visual futures through Scale-wise autoregressive Scaling, enabling action-controllable and long-horizon robotic simulation, and consistently benefits from on-policy alignment in reducing action drift and long-horizon inconsistency.
Maeve Zhang, Rainy Sun, Xiang Wang et al.· 2 citations· ⚡1
This work presents LeFlow, which learns a reusable latent trajectory prior operating directly in the latent dynamics space from the world model, and argues that latent world models should support not only prediction but reusable planning priors.
Hsiang-Wei Huang, Jian-Xu Shangguan, Junbin Lu et al.· 5 citations
Long-horizon world-model planning typically relies on autoregressive rollouts, where predicted states are repeatedly fed back into the model. This preserves temporal structure but creates a horizon-length sequential path and exposes later predictions to recursive decoded-state feedback. We introduce Parallel Predictive...
Wan-Jin Feng, Bao-Bing Zhang, Ao Yu et al.· 0 citations
This work introduces Latent Energy Action Planning (LEAP), which treats the complete action horizon as a differentiable variable and optimizes it through a frozen LeWorldModel (LeWM) through a frozen LeWorldModel representation and predictor.