⌁ Awesome World Model Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

Value-guided action planning with JEPA world models

World Modeling Workshop 26 2026 46.9 method

TLDR

Enhances JEPA world models for action planning by shaping representation space to approximate value functions, improving performance on simple control tasks.

Reasoning

The paper presents a novel method to integrate value-guided planning into JEPA world models, which is a clear strength. However, the evaluation is limited to simple control tasks, and there is no evidence of real-world applicability or scalability to complex environments.

Read-first score

Read-first score 46.9, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 27.

Recency 8%
100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 25%
60

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=baseline

Topical relevance 42%
38.6

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 25%
30

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 464.

Keyword Scores

world model
9
model-based reinforcement learning world model
6
world dynamics prediction
5
interactive world model
3
generative world model
2
world simulator
1
video world model
1

Deep Analysis

Innovations

  • Shaping the representation space of JEPA world models so that the negative goal-conditioned value function for a reaching cost is approximated by a distance (or quasi-distance) between state embeddings.
  • A practical method to enforce this constraint during training of JEPA models.
  • Demonstration of significantly improved planning performance compared to standard JEPA models on simple control tasks.

Methodology

The approach builds on Joint-Embedded Predictive Architectures (JEPA) and modifies the training objective to enforce that the distance between state embeddings approximates the negative goal-conditioned value function for a reaching cost. The method is evaluated on simple control tasks by comparing planning performance against standard JEPA baselines.

Key Results

The proposed value-guided planning with JEPA world models leads to significantly improved planning performance compared to standard JEPA models on simple control tasks.

Limitations

  • Only evaluated on simple control tasks, leaving generalization to more complex environments unaddressed.
  • The method assumes a reaching cost structure for the goal-conditioned value function, which may not hold for all tasks.

Resources

Paper

Tags