# Depth-Regularized JEPA World Models Learn More Transferable Representations from Real Outdoor Robot Data

arXiv: https://arxiv.org/abs/2607.16314 · HTML: https://mlsanity.com/paper/2607.16314/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Latent predictive world models, one of 11 directions in World models.
- Builds on 20 of its 21 references that are in the map.
- First posted on arXiv 15 Jul 2026.

## Placed in

- World models / [Latent predictive world models](https://mlsanity.com/wm/latent-predictive-world-models/)

## Builds on (most cited first, 20 of 20)

- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Dream to Control: Learning Behaviors by Latent Imagination](https://mlsanity.com/paper/1912.01603/) (2019)
- [World Models](https://mlsanity.com/paper/1803.10122/) (2018)
- [Learning Latent Dynamics for Planning from Pixels](https://mlsanity.com/paper/1811.04551/) (2018)
- [Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture](https://mlsanity.com/paper/2301.08243/) (2023)
- [V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning](https://mlsanity.com/paper/2506.09985/) (2025)
- [Revisiting Feature Prediction for Learning Visual Representations from Video](https://mlsanity.com/paper/2404.08471/) (2024)
- [RepVGG: Making VGG-style ConvNets Great Again](https://mlsanity.com/paper/2101.03697/) (2021)
- [DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning](https://mlsanity.com/paper/2411.04983/) (2024)
- [LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels](https://mlsanity.com/paper/2603.19312/) (2026)
- [Understanding self-supervised Learning Dynamics without Contrastive Pairs](https://mlsanity.com/paper/2102.06810/) (2021)
- [Enhancing End-to-End Autonomous Driving with Latent World Model](https://mlsanity.com/paper/2406.08481/) (2024)
- [MDP Homomorphic Networks: Group Symmetries in Reinforcement Learning](https://mlsanity.com/paper/2006.16908/) (2020)
- [MobileOne: An Improved One millisecond Mobile Backbone](https://mlsanity.com/paper/2206.04040/) (2022)
- [World4Drive: End-to-End Autonomous Driving via Intention-aware Physical Latent World Model](https://mlsanity.com/paper/2507.00603/) (2025)
- [Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models](https://mlsanity.com/paper/2502.14819/) (2025)
- [TartanGround: A Large-Scale Dataset for Ground Robot Perception and Navigation](https://mlsanity.com/paper/2505.10696/) (2025)
- [seq-JEPA: Autoregressive Predictive Learning of Invariant-Equivariant World Models](https://mlsanity.com/paper/2505.03176/) (2025)
- [A Mixed Diet Makes DINO An Omnivorous Vision Encoder](https://mlsanity.com/paper/2602.24181/) (2026)

## Built on by (most cited first, 2 of 2)

- [Spectral-Target Physical Latent Structuring for JEPA-Style World Models](https://mlsanity.com/paper/2609.04264/) (2026)
- [Underwater C3-JEPA: An Object-Centric Cross-View World Model for ROV Salvage](https://mlsanity.com/paper/2609.30214/) (2026)

## Abstract (arXiv)

World models, especially based on JEPA architectures, have been shown to learn robust dynamics of various environments. However, learning from visually complex real-world data remains a challenge, especially in unpredictable outdoor environments. We introduce depth as a geometric prior during training in learning more robust latent dynamics directly from robot video data and handling visual complexity. This combines depth supervision with an isotropy-inducing latent regularizer (SIGReg), maximizing task-agnostic latent diversity while constraining how that diversity is organized, with the combined objective targeting the highest-entropy representation consistent with scene geometry. To satisfy this greater complexity without increasing inference time, we also add training-only overparameterization. Training an 18M-parameter model on video from a real agricultural robot, we evaluate with frozen-representation visual odometry probes, predictor-based surprise detection, and multi-step latent rollout fidelity. Compared to the baseline LeWM, our method lowers visual odometry probe error by 33%, substantially increases surprise-score separation both in-domain and on the out-of-domain TartanGround benchmark, and improves multi-step rollout fidelity under domain shift, with gains that grow with rollout horizon. Notably, we also see improvements in surprise-score separation on physics understanding that is not directly tied to 3D geometry, such as lighting and shadows. These results show that a lightweight training-time geometric prior makes a compact JEPA world model more useful and more transferable on real outdoor data with strong underlying representations, without adding inference overhead. Our work suggests that depth as a physically grounded prior can enhance world model generalization on a variety of tasks.