D-JEPA: A Decision-Aligned Latent World Model

原文：D-JEPA: A Decision-Aligned Latent World Model

作者：Shuaijun Liu, Chengyu Wu, Qifu Wen, Feiyang You, Chenglong Zhang, Shuyang Hao, Xi Lin, Ningxin Su

来源：arXiv cs.LG（机器学习）

正文

Computer Science > Robotics

arXiv:2609.24749v1 (cs)

[Submitted on 21 Sep 2026]

Title:D-JEPA: A Decision-Aligned Latent World Model

Authors:Shuaijun Liu, Chengyu Wu, Qifu Wen, Feiyang You, Chenglong Zhang, Shuyang Hao, Xi Lin, Ningxin Su

View a PDF of the paper titled D-JEPA: A Decision-Aligned Latent World Model, by Shuaijun Liu and 7 other authors

View PDF

HTML (experimental)

Abstract:Latent world models predict the consequences of actions, but accurate prediction does not guarantee that latent distance reflects which candidate will execute successfully. We identify a decision-local prediction gap: among the few futures competing for execution, a candidate predicted closer to the goal can produce a worse realized outcome than an available alternative. We introduce D-JEPA, a decision-aligned latent world model that learns decision-relevant relations among candidate futures from executed outcomes. A bounded, permutation-equivariant operator jointly reasons over goal-relative predictive features and ordinal evidence, refining pretrained predictive geometry where action choices are most consequential. Restricted predictor adaptation and a shared ordinal interface extend this alignment across complementary predictive geometries. D-JEPA further realizes the learned decision structure in JEPA-compatible future representations, enabling deployment through native latent-distance planning. Evaluations across latent control, manipulation, pretrained action-producing models, physical robots and autonomous driving demonstrate improved action selection, including 87.89% success on PushT, a 15.04-point average gain on RoboTwin, and a 17-point gain on physical robot tasks. These results establish decision-relevant relational structure as a direct bridge between predictive world modeling and effective control.

Comments:

26 pages, including references and appendices. Project website: this https URL

Subjects:

Robotics (cs.RO); Machine Learning (cs.LG)

Cite as:

arXiv:2609.24749 [cs.RO]

(or

arXiv:2609.24749v1 [cs.RO] for this version)

https://doi.org/10.48550/arXiv.2609.24749

Focus to learn more

arXiv-issued DOI via DataCite (pending registration)

主题

机器人 · 机器学习

由「前沿雷达」于 2026-09-21 采集。正文取自原文页面，已保留出处链接。标题与正文版权归原作者所有。