mlsanityData to 14 Oct 2026

How Long, Not How Close: A Learned Temporal Metric for Planning in Latent World Models

arXiv 2610.04988 ↗First posted 4 Oct 2026cs.LGcited by 0

On the map: World models → Latent predictive world models

Abstract (arXiv)

Latent world models plan by rolling a frozen predictor forward under candidate action sequences and ranking the candidates by the latent distance between their imagined end state and the goal. However, this ranking breaks down when the goal lies several plans away, because the latent distance measures how closely an end state resembles the goal rather than how far it remains from reaching it. To address this, we propose TEMPO, a temporal-distance planning objective that leaves the world model untouched, learns only from the recorded trajectories already used to train it, and adds negligible cost to the planner's search. TEMPO learns a small map of the frozen latent in which the distance between two states of an episode reflects the number of environment steps between them, and blends this distance into the planner's cost. It requires no rewards, policies or success labels and, being a cost rather than a model, applies to frozen world models with one latent vector per state that plan by a latent distance. We evaluate TEMPO on eleven simulated environments (e.g., maze navigation, tabletop pushing, robotic arm control and three-dimensional manipulation) with the LeWM and PLDM planners. With a small MLP that adds at most 0.3% to a plan's arithmetic, TEMPO improves both planners at every goal distance, including the one-plan setting of their evaluations, raises LeWM from 36% to 99% on TwoRoom three plans from the goal, and remains competitive on a broad range of 2D and 3D navigation, reaching and manipulation tasks.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Latent predictive world models, one of 11 directions in World models. Builds on 32 of its 42 references that are in the map. First posted on arXiv 4 Oct 2026.

Builds on

32 of 42 references.

Closest ancestors · 11

Same direction, newest first

  1. 2026Temporal-Distance JEPA: Plan-Aware Representation Learning for Latent World Model Predictive Control28 Jul
  2. 2026SAGE: Subgoal-Conditioned Action Generation for Latent World Model Planning20 Jul
  3. 2026AdaJEPA: An Adaptive Latent World Model30 Jun
  4. 2026Predictive but Not Plannable: RC-aux for Latent World Models8 May
  5. 2026Hierarchical Planning with Latent World Models3 Apr
  6. 2026LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels13 Mar
  7. 2024Revisiting Feature Prediction for Learning Visual Representations from Video15 Feb
  8. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture19 Jan
  9. 2022Deep Hierarchical Planning from Pixels8 Jun
  10. 2022Temporal Difference Learning for Model Predictive Control9 Mar
  11. 2018Learning Latent Dynamics for Planning from Pixels12 Nov

Other references · 21

  1. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  2. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,653
  3. 2020D4RL: Datasets for Deep Data-Driven Reinforcement Learningcited by 1,427
  4. 2021Offline Reinforcement Learning with Implicit Q-Learningcited by 1,293
  5. 2021VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learningcited by 1,057
  6. 2018DeepMind Control Suitecited by 956
  7. 2018Recurrent World Models Facilitate Policy Evolutioncited by 785
  8. 2023TD-MPC2: Scalable, Robust World Models for Continuous Controlcited by 512
  9. 2024RoboCasa: Large-Scale Simulation of Everyday Tasks for Generalist Robotscited by 407
  10. 2022Transformers are Sample-Efficient World Modelscited by 328
  11. 2024Diffusion for World Modeling: Visual Details Matter in Ataricited by 300
  12. 2025LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristicscited by 254
9 more
  1. 2024OGBench: Benchmarking Offline Goal-Conditioned RLcited by 238
  2. 2019Planning with Goal-Conditioned Policiescited by 149
  3. 2023HIQL: Offline Goal-Conditioned RL with Latent States as Actionscited by 141
  4. 2023Optimal Goal-Reaching Reinforcement Learning via Quasimetric Learningcited by 121
  5. 2020Sample-efficient Cross-Entropy Method for Real-time Planningcited by 95
  6. 2024Foundation Policies with Hilbert Representationscited by 94
  7. 2025Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Modelscited by 89
  8. 2020Long-Horizon Visual Planning with Goal-Conditioned Hierarchical Predictorscited by 62
  9. 2026stable-worldmodel: A Platform for Reproducible World Modeling Research and Evaluationcited by 20

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Beyond a single latent space: a dual-latent world model for long-horizon planning
  2. 2026Beyond the Next Step: Variable-Length Latent World Models for Long-Horizon Planning
  3. 2026Latent Goal Prediction from Language for Model-Based Planning
  4. 2026The Planning Limits of Latent World Models
  5. 2026HaM-World: Soft-Hamiltonian World Models with Selective Memory for Planning
  6. 2026The Objective Is the Bottleneck: Latent World Models Encode What Their Planners Cannot Use
  7. 2020World Model as a Graph: Learning Latent Landmarks for Planning
  8. 2026IMWM: Intuition Models Complement World Models for Latent Planning