Skip to content

DeepJEPA: Scaling World Models from Within

Sep 2026 · 1 citation · 59 references
Computer Science

TL;DR

DeepJEPA is introduced, a weight-tied joint-embedding predictive world model that treats transition depth as an inner test-time scaling axis and learns when another recurrent update is worth computing for each candidate and rollout step.

Abstract

World-model planners typically scale outward by rolling farther, sampling more trajectories, or optimizing longer, while assigning the same computation to every imagined transition. We show that making every transition uniformly deeper wastes computation and can degrade planning because useful refinement is concentrated at a small set of decision-critical events. We introduce DeepJEPA, a weight-tied joint-embedding predictive world model that treats transition depth as an inner test-time scaling axis and learns when another recurrent update is worth computing for each candidate and rollout step. Across five visual-control settings, DeepJEPA improves or matches the strongest fixed-depth planner while averaging only 1.00-1.26 updates per transition. Its additional computation concentrates at contact onset and sustained object interaction, where latent corrections can change which candidates enter the planner's elite set and which action is selected. Representation probes further show that improved planning does not require uniformly better object-state decodability. DeepJEPA therefore reframes world-model scaling as a problem of allocating internal computation where it can change the planner's decision: think deeper at decision-critical transitions instead of making every rollout uniformly deeper or longer.

View source

Similar papers

#artificial intelligence Preprint Sep 2026

LePlanner: An Iterative Amortized Controller For World Models

LePlanner is an amortized iterative controller that learns to construct and refine latent action sequences through a frozen world-model predictor, enabling fast, horizon-aware, nonlinear physical control without online optimization.

Saksham Bansal, Om Naphade, Chayan Aggarwal et al. · 1 citation · ⚡1
Preprint Oct 2026

Sampling-based Kinodynamic Planning with World Models

Visual world models predict the consequences of actions from images, but planning with them is typically done by optimizers such as CEM and MPPI. However, these optimizers stall on long-horizon tasks where progress requires motion the objective does not reward, such as moving around a block before pushing it. This pape...

Yi-Tian Gao, Han-Lan Yang, Zachary Kingston · 0 citations
Preprint Nov 2026

Time-to-Goal: Learned Temporal Metrics for Long-Horizon Planning with JEPA World Models

Joint-Embedding Predictive Architectures (JEPAs) support planning by comparing predicted latent states with an encoded goal. For this comparison to guide temporal progress, states closer in time to the goal should receive lower costs. However, training representations for prediction and distribution regularization does...

Rui-Long Yu, Fei Ye · 0 citations

Aim Short to Reach Far: Your Frozen World Model Can Plan Better Than You Think

This work introduces Anchored Planning, which retrieves a recorded segment whose start and end resemble the current and goal observations, then aims at an observation shortly after its start, and without additional training, planning toward observed targets outperforms the LeWM planner on every task in the authors' lon...

Xvyuan Liu, Jian-Jie Fang, Wei Wu et al. · 0 citations
#machine learning Preprint Sep 2026

Learning to Plan from Random Exploration

Random exploration reveals how an environment can be traversed before a goal is specified. Can this experience support long-range planning without policy-improvement training? Our random-walk analysis explains what temporal relations contain: short horizons reveal geodesic geometry in the diffusion limit, while longer...

De-Qian Kong, Guang-Yan Sun, Sheng Cheng et al. · 0 citations
#machine learning Preprint Oct 2026

H-JEPA: End-to-End Learning of Hierarchical World Models for Visual Planning

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

Related blog posts

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.