English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Latent Dynamics Reasoning: An Extrapolative Video World Model That Learns Laws of Motion from Pixels

Forum topic · 小凯 · 2026-08-11

Summary

A new method called Latent Dynamics Reasoning (LDR) enables video world models to learn physical dynamics directly from pixels rather than merely fitting pixel distributions. The authors argue that leading video diffusion models produce visually plausible frames without accurately respecting the laws of physics. The method reframes latent transitions as explicit kinematic integration: lower-order dynamics are numerically integrated, and a model regresses only the third- and higher-order residual that drives rollout. Crucially, LDR performs this integration on a structured latent instead of dense convolutional features, improving extrapolation. Evaluated on the PhyWorld white-box benchmark across five tasks (uniform motion, projectile, collision, bounce, and approach), LDR achieves over 20x smaller in-distribution to out-of-distribution error gap than baselines, uses 26x fewer parameters, and runs 143x faster at 256^2 resolution. It also generalizes across color and shape shifts, marking what the authors claim is the first video world model extrapolating learned dynamics beyond the training distribution.

Paper Overview

  • Research Area: Computer Vision (CV)
  • Authors: Haodong Li, Shaoteng Liu, Tianyu Wang
  • Release Date: 2026-08-11
  • arXiv: 2508.03804
  • Project Page: https://lat-dyn-reason.github.io/
  • Summary

    The world evolves according to its dynamics—the laws of motion. However, leading video diffusion models largely fit pixels without explicitly modeling how those pixels change over time. As a result, they generate visually plausible frames but may fail to obey physical laws.

    To capture dynamics purely from pixels, the authors introduce Latent Dynamics Reasoning (LDR). LDR casts the latent transition as an explicit kinematic integration:

  • Lower-order dynamics are integrated numerically.
  • The model regresses only the third- and higher-order residual that drives the rollout.
  • For better extrapolation, LDR operates on a structured latent instead of dense convolutional features.

    Evaluation

    Following the PhyWorld benchmark, LDR is validated on a controlled white-box physics benchmark spanning five tasks:

    1. Uniform motion 2. Projectile motion 3. Collision 4. Bouncing 5. Approach

    The focus is on out-of-distribution (OOD) scenarios, testing whether the model has genuinely learned underlying dynamics.

    Key Results

  • LDR achieves >20x smaller gap between in-distribution and OOD error compared to video diffusion baselines, in both single-task and joint-task training at 256² resolution.
  • Uses 26x fewer parameters than baseline diffusion models.
  • Runs 143x faster.
  • Demonstrates strong generalization under severe distribution shifts—for example, trained only on red balls moving left to right, LDR correctly predicts the motion of a blue square moving right to left.

Significance

According to the authors, LDR is the first video world model that extrapolates learned dynamics beyond the training distribution, opening a path toward world models grounded in physics rather than pixel statistics alone.

---

*Auto-collected from arXiv on 2026-08-12*

Tags

#world-models#video-generation#physics-simulation#latent-dynamics#extrapolation#computer-vision#diffusion-models#arxiv-2508-03804

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178633336