mlsanityData to 14 Oct 2026

World Model Control by Trajectory Reachability Metrics

arXiv 2605.22164 ↗First posted 21 May 2026cs.LG cs.ROcited by 8

On the map: World models → Latent predictive world models

Abstract (arXiv)

Latent world models can learn representations that contain information needed for control, while the downstream controller may still rank candidate actions poorly when it relies on terminal latent distance alone. We study this failure in a fixed encoder and introduce trajectory reachability metrics (TRM), a small temporal pairwise cost trained from logged trajectories and used to rank predicted endpoints of a candidate action sequence against a goal. In the TwoRoom evaluation on 100 episodes from a high distance range, the original controller with LeWorldModel reaches 7.0% mean success. Temporal TRM trained after excluding all evaluation episodes reaches 96.7%, shuffled label controls stay at 0.0%, and TRM also improves PLDM from 32.7% to 84.0%. TopoNav, a separate pixel navigation task, shows the same link between repaired ranking and closed loop control. The selection audit with shared candidates (SASC) and rowspace interventions show that the gain comes from reweighting latent directions carrying the action decision. The XY rowspace contributes less than 1% of terminal latent MSE but carries most of the information needed for control. Coverage and boundary tests define the scope. Balanced doorway coverage restores a 0.0% coverage failure to 100.0% success, while an unseen wall orientation and contact rich PushT expose layout, dynamics, and recovery limits.

Key facts. Cited by 8 papers in the map (data through 14 Oct 2026). Placed in Latent predictive world models, one of 11 directions in World models. Builds on 15 of its 41 references that are in the map. First posted on arXiv 21 May 2026.

Builds on

15 of 41 references.

Closest ancestors · 3

Same direction, newest first

  1. 2026Predictive but Not Plannable: RC-aux for Latent World Models8 May
  2. 2026LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels13 Mar
  3. 2025Closing the Train-Test Gap in World Models for Gradient-Based Planning10 Dec

Other references · 12

  1. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,838
  2. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,653
  3. 2018World Modelscited by 1,513
  4. 2017Hindsight Experience Replaycited by 1,136
  5. 2018Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Modelscited by 774
  6. 2019When to Trust Your Model: Model-Based Policy Optimizationcited by 773
  7. 2018Data-Efficient Hierarchical Reinforcement Learningcited by 455
  8. 2018Visual Reinforcement Learning with Imagined Goalscited by 251
  9. 2018Visual Foresight: Model-Based Deep Reinforcement Learning for Vision-Based Robotic Controlcited by 230
  10. 2019Search on the Replay Buffer: Bridging Planning and Reinforcement Learningcited by 194
  11. 2026stable-worldmodel: A Platform for Reproducible World Modeling Research and Evaluationcited by 20
  12. 2023Goal-Conditioned Reinforcement Learning with Disentanglement-based Reachability Planningcited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 8.

8 papers.

  1. 2026ATM: Action-Consistency Transfer Matrix for Diagnosing and Improving Latent World Modelscited by 3
  2. 2026Action from Adjacent Set in Physical Space Outperforms the Best Prediction in World Modelscited by 2
  3. 2026Reinforced Planning with Latent World Modelscited by 2
  4. 2026The Evaluation Protocol Determines the Result: An Independent Reproduction of LeWorldModel on TwoRoomcited by 0
  5. 2026ARC-Bench: Closed-Loop Replanning Masks Broken Action Ranking in Frozen JEPA World Modelscited by 0
  6. 2026Anisotropic Representations Improve Planning in JEPA World Modelscited by 0
  7. 2026Supervise What Decides Success: Criterion-Aligned Auxiliary Losses for Latent World-Model Planningcited by 0
  8. 2026When Low Prediction Error Misleads Planning: Diagnosing Representation, Dynamics, and Decision Failures in Latent World Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Traj-LeWM: Path-Aware World-Model Planning via Latent Trajectory Cost
  2. 2025Latent Action World Models for Control with Unlabeled Trajectories
  3. 2026Directed Temporal Representations for Offline Visual Control
  4. 2026Decision-Metric Alignment in Latent World Models: Diagnostics and Action-Conditioned Objectives for MPC Planning
  5. 2026WALT: Learning World-Model-Aligned Latent Trajectories for Autonomous Driving
  6. 2026How Long, Not How Close: A Learned Temporal Metric for Planning in Latent World Models
  7. 2026Beyond a single latent space: a dual-latent world model for long-horizon planning
  8. 2026Predicting Consequences and Reinforcing Navigation Policies with Latent World Models