RoboJEPA: Scaling Laws for Robotic Latent World Models (8B-Parameter JEPA)

Paper Overview

Field: Machine Learning Authors: Artem Zholus, Nicolas Beltran-Velez, Jianhao Yuan, Sarath Chandar, Tushar Nagarajan, Daniel Severo, Koustuv Sinha, Michal Drozdzal, Adriana Romero Soriano, Jeannette Bohg, Nicolas Ballas, Mahmoud Assran arXiv: 2610.10515

Abstract

Latent world models have shown a remarkable ability to predict future states and to plan in the real world. In practice, however, we lack a principled way to estimate how their capabilities scale with model size, data, and compute—an open problem that slows progress in the field.

In this work the authors present RoboJEPA, a world model based on the Joint Embedding Predictive Architecture (JEPA), trained on a large-scale dataset spanning 12 robotic embodiments. Key findings:

• RoboJEPA's imagination error (the error of its latent rollouts) follows a second-order power law in compute, allowing prediction of model quality well beyond the scale at which the law is fit.
• Downstream robotic planning performance improves predictably with compute, and imagination error is strongly correlated with it, making imagination error a reliable proxy for real-robot evaluation.
• Latent world models can be deployed zero-shot as robotic agents, planning toward a single goal image to complete long-horizon tasks on real hardware.

All model checkpoints, training code, and robot deployment code are released. To the authors' knowledge, this is the first work establishing scaling laws for multi-embodiment robotic world models on real robot data; RoboJEPA, at 8B parameters, is the largest JEPA predictive model to date.

--- *Source: zhichai.net forum post, collected 2026-10-09.*