OpenTrain AI
No verified implementation yet

Co-Evolving Latent Action World Models

Yucen Wang, Fengming Zhang, De-Chuan Zhan, Li Zhao, Kaixin Wang +1 more

October 30, 2025arXiv: 2510.26433
0 repos~a few days to reproduce
arXiv PDF

Abstract

Adapting pretrained video generation models into controllable world models via latent actions is a promising step towards creating generalist world models. The dominant paradigm adopts a two-stage approach that trains latent action model (LAM) and the world model separately, resulting in redundant training and limiting their potential for co-adaptation. A conceptually simple and appealing idea is to directly replace...

Results & Benchmarks

TaskDatasetMetricValue
Scientific computingRoboDesk on whole VP 2Push Drawer4.00

Hardware Requirements

  • Expect multi-day setup/compute for meaningful reproduction based on current guidance.

Best Implementation

Maintained implementation evidence is not confirmed for this paper yet.

Use the Implementation Status and Reproduction Path sections below for the current action plan.

Reproduction Path

Follow this baseline workflow to decide if this paper is worth immediate prototyping.

  1. 1

    Use the paper and benchmark evidence to scope a baseline reproduction plan.

  2. 2

    Track assumptions and missing details in an experiment log before coding.

Time to first repro: a few daysEstimate is based on paper-only reproduction flow

Additional Implementations

No additional verified repositories beyond the primary recommendation.

Hugging Face Artifacts

No trustworthy direct or curated related Hugging Face artifacts were found yet.

Research Context