# What Makes Video World Model Latents Action-Relevant: Prediction over Reconstruction

arXiv: https://arxiv.org/abs/2606.07687 · HTML: https://mlsanity.com/paper/2606.07687/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Placed in Video-generation world models and simulators, one of 11 directions in World models.
- Builds on 31 of its 34 references that are in the map.
- First posted on arXiv 5 Jun 2026.

## Placed in

- World models / [Video-generation world models and simulators](https://mlsanity.com/wm/video-generation-world-models-and-simulators/)

## Builds on (most cited first, 30 of 31)

- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control](https://mlsanity.com/paper/2307.15818/) (2023)
- [OpenVLA: An Open-Source Vision-Language-Action Model](https://mlsanity.com/paper/2406.09246/) (2024)
- [π0: A Vision-Language-Action Flow Model for General Robot Control](https://mlsanity.com/paper/2410.24164/) (2024)
- [LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning](https://mlsanity.com/paper/2306.03310/) (2023)
- [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://mlsanity.com/paper/2203.12602/) (2022)
- [Mastering Diverse Domains through World Models](https://mlsanity.com/paper/2301.04104/) (2023)
- [Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning](https://mlsanity.com/paper/1910.10897/) (2019)
- [Curiosity-driven Exploration by Self-supervised Prediction](https://mlsanity.com/paper/1705.05363/) (2017)
- [SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features](https://mlsanity.com/paper/2502.14786/) (2025)
- [Cosmos World Foundation Model Platform for Physical AI](https://mlsanity.com/paper/2501.03575/) (2025)
- [V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning](https://mlsanity.com/paper/2506.09985/) (2025)
- [CALVIN: A Benchmark for Language-Conditioned Policy Learning for Long-Horizon Robot Manipulation Tasks](https://mlsanity.com/paper/2112.03227/) (2021)
- [DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning](https://mlsanity.com/paper/2411.04983/) (2024)
- [Transformers are Sample-Efficient World Models](https://mlsanity.com/paper/2209.00588/) (2022)
- [Latent Action Pretraining from Videos](https://mlsanity.com/paper/2410.11758/) (2024)
- [LIBERO-PRO: Towards Robust and Fair Evaluation of Vision-Language-Action Models Beyond Memorization](https://mlsanity.com/paper/2510.03827/) (2025)
- [Learning to Poke by Poking: Experiential Learning of Intuitive Physics](https://mlsanity.com/paper/1606.07419/) (2016)
- [Training Agents Inside of Scalable World Models](https://mlsanity.com/paper/2509.24527/) (2025)
- [WorldModelBench: Judging Video Generation Models As World Models](https://mlsanity.com/paper/2502.20694/) (2025)
- [WorldArena: A Unified Benchmark for Evaluating Perception and Functional Utility of Embodied World Models](https://mlsanity.com/paper/2602.08971/) (2026)
- [Loss is its own Reward: Self-Supervision for Reinforcement Learning](https://mlsanity.com/paper/1612.07307/) (2016)
- [Scaling Language-Free Visual Representation Learning](https://mlsanity.com/paper/2504.01017/) (2025)
- [EWMBench: Evaluating Scene, Motion, and Semantic Quality in Embodied World Models](https://mlsanity.com/paper/2505.09694/) (2025)
- [World-in-World: World Models in a Closed-Loop World](https://mlsanity.com/paper/2510.18135/) (2025)
- [DynaMo: In-Domain Dynamics Pretraining for Visuo-Motor Control](https://mlsanity.com/paper/2409.12192/) (2024)
- [R-Bench: Are your Large Multimodal Model Robust to Real-world Corruptions?](https://mlsanity.com/paper/2410.05474/) (2024)
- [A Control-Centric Benchmark for Video Prediction](https://mlsanity.com/paper/2304.13723/) (2023)
- [Interpreting Physics in Video World Models](https://mlsanity.com/paper/2602.07050/) (2026)
- [Reconstruction or Semantics? What Makes a Latent Space Useful for Robotic World Models](https://mlsanity.com/paper/2605.06388/) (2026)

## Built on by (most cited first, 9 of 9)

- [INTACT: Isomorphic Intent-to-Action Learning for Search-Free World Models](https://mlsanity.com/paper/2607.26056/) (2026)
- [Diagnosing JEPA World Models with Action-Conditioned Predictive Consistency](https://mlsanity.com/paper/2608.12939/) (2026)
- [Low-Rank Dynamics-Effective Latent Carriers for Counterfactual Rollout in Learned World Models](https://mlsanity.com/paper/2608.15156/) (2026)
- [Improving Cross-embodiment Transfer in Latent Action Models with Action-Similarity Supervision](https://mlsanity.com/paper/2609.19846/) (2026)
- [Rethinking Representations for World-Action Modeling](https://mlsanity.com/paper/2609.38163/) (2026)
- [From Generation to Simulation: How Far Are World Models from Being True Simulators?](https://mlsanity.com/paper/2608.23070/) (2026)
- [Frozen Flows Forget: Diagnosing and Restoring Lost Motion in a Latent-flow World Model](https://mlsanity.com/paper/2609.28414/) (2026)
- [The Planning Limits of Latent World Models](https://mlsanity.com/paper/2609.39235/) (2026)
- [Keeping JEPA World Models Plannable When Little of the Frame Moves](https://mlsanity.com/paper/2610.03137/) (2026)

## Abstract (arXiv)

Video world models are increasingly used to provide predictive visual representations, yet it remains unclear which pretraining signals induce action-relevant structure in their latent spaces. We study this question through a unified probe-based evaluation across diverse encoder families, including image-only self-supervision, video pretraining with and without latent prediction, reconstruction-based autoencoders, diffusion models, and shortcut-forcing dynamics models. Using a common inverse-dynamics probing objective, we find that action-relevant structure is driven primarily by temporal video pretraining rather than pixel reconstruction fidelity: models with strong pixel decoding quality can exhibit near-zero action recoverability, while video-pretrained self-supervised encoders consistently achieve the best Pareto trade-off between visual fidelity and action prediction. Comparing V-JEPA and VideoMAE further shows that most gains arise from natural-video temporal context, with feature-level latent prediction providing a smaller additional benefit. These trends transfer across robotic benchmarks, though CALVIN reveals that static-environment tasks can partially mask the importance of temporal structure by allowing strong image priors to suffice. Finally, inverse-dynamics supervision substantially improves robustness to visual corruption, suggesting that action-aware objectives regularize latent geometry beyond clean-setting performance. Our results identify temporal predictive structure -- not reconstruction fidelity -- as the primary ingredient underlying action-relevant video representations.