Frozen in a Frame: The Velocity Blind Spot in JEPA World Models

arXiv:2610.04585 · cs.CV · Submitted 2026-10-03 · Read on arXiv

cs.CV

Submitted: 2026-10-03

Updated: 2026-10-03

Comments: 32 pages, 17 figures, 12 tables. Code, model checkpoints, and project page are available via links in the paper

Code: https://github.com/OliverZ-dot/TI-JEPA

Project page: https://oliverz-dot.github.io/ti-jepa-homepage

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• World Model — 1× in this paper · explained in 7 episodes

Sources

• Revisiting Feature Prediction for Learning Visual Representations from Video
• Fast LeWorldModel
• World Models
• Mastering Diverse Domains through World Models
• When Does LeJEPA Learn a World Model?
• LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels
• Is Forward Prediction Enough? Physical State Grounding for JEPA World Models

Related papers

• Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
• AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
• Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
• MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
• TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
• A Survey on Efficient Vision-Language-Action Models