aiqbench analysis
This research introduces SAGE, a planning framework that leverages latent world models by using subgoals as structured priors to guide action generation. By decomposing long-horizon tasks into manageable semantic steps, the method significantly boosts success rates in robotic manipulation benchmarks like PushT and OGBench Cube without sacrificing short-term performance. The approach addresses a critical bottleneck where fixed candidate budgets fail to explore sufficiently large state spaces effectively over extended planning horizons.
SAGE: Subgoal-Conditioned Action Generation for Latent World Model Planning
arXiv:2607.17973v1 Announce Type: new Abstract: Latent world models have emerged as a powerful planning paradigm by learning action-conditioned predictive dynamics and using them as internal simulators to imagine and evaluate candidate action sequences. However, as the planning horizon grows, performance becomes increasingly constrained by proposal quality: a fixed candidate budget must search an exponentially larger action space, making it difficult to expose the world model to high-quality candidate futures for evaluation. In this paper, we introduce a prior-conditioned planner that replaces random proposal initialization with structured guidance. At each planning stage, a goal-conditioned generator predicts the next reachable latent subgoal for a specified duration, which is then used to condition the generation of candidate action sequences. To capture semantic information across temporal scales, we use subgoals of varying durations as priors, balancing fine-grained local control with higher-level long-horizon progress. Then the frozen world model evaluates and refines these subgoal-conditioned proposals before execution. Experiments on PushT and OGBench Cube show that coupling latent subgoal decomposition with prior-conditioned action generation substantially improves long-horizon planning while preserving strong short-horizon performance. To be specific, when the target offset is $150$, it raises PushT success from $12.7\%$ to $64.7\%$ and OGBench Cube success from $26.7\%$ to $67.3\%$.
Abstract:Latent world models have emerged as a powerful planning paradigm by learning action-conditioned predictive dynamics and using them as internal simulators to imagine and evaluate candidate action sequences. However, as the planning horizon grows, performance becomes increasingly constrained by proposal quality: a fixed candidate budget must search an exponentially larger action space, making it difficult to expose the world model to high-quality candidate futures for evaluation. In this paper, we introduce a prior-conditioned planner that replaces random proposal initialization with structured guidance. At each planning stage, a goal-conditioned generator predicts the next reachable latent subgoal for a specified duration, which is then used to condition the generation of candidate action sequences. To capture semantic information across temporal scales, we use subgoals of varying durations as priors, balancing fine-grained local control with higher-level long-horizon progress. Then the frozen world model evaluates and refines these subgoal-conditioned proposals before execution. Experiments on PushT and OGBench Cube show that coupling latent subgoal decomposition with prior-conditioned action generation substantially improves long-horizon planning while preserving strong short-horizon performance. To be specific, when the target offset is $150$, it raises PushT success from $12.7\%$ to $64.7\%$ and OGBench Cube success from $26.7\%$ to $67.3\%$.
Artificial Intelligence (cs.AI)
arXiv:2607.17973 [cs.AI]
(or arXiv:2607.17973v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2607.17973 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 · 291 words
Related coverage
LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies
arXiv · Jun 16, 2026
Latent Goal Prediction from Language for Model-Based Planning
arXiv · Oct 6, 2026
Beyond a single latent space: a dual-latent world model for long-horizon planning
arXiv · Sep 30, 2026
SUAVE: Unified Video-Action Models via Masked Diffusion
arXiv · Oct 6, 2026