ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?

Hugging Face Daily Papers Papers
image-editing world-action-models robot-control flow-matching computer-vision robotics efficient-inference

Summary

ImageWAM proposes replacing video generation with pretrained image editing models in world action models for robot control, achieving superior performance while reducing FLOPs to 1/6 and latency to 1/4 of video-based approaches.

World Action Models (WAMs) commonly rely on video generation to bridge visual world modeling and robot control. However, video-based WAMs face three coupled limitations: dense multi-frame future tokens make inference costly, full video prediction spends capacity on action-irrelevant temporal and appearance details, and long-horizon future imagination may introduce errors that mislead action prediction. These issues raise a simple question: Does world action model really need video generation? We propose ImageWAM, a simple WAM framework that repurposes pretrained image editing models for robot action prediction. In contrast to video generation, image editing provides a better-matched prior: it only needs to model a target-frame transformation, focuses on action-relevant current-to-target visual differences, and grounds task instructions to localized visual changes through edit pretraining. In practice, ImageWAM does not decode the target frame at inference time; instead, it conditions a flow-matching action expert on the KV caches produced by image-editing denoising, using them as a compact world-action context. ImageWAM outperforms standard VLA baselines and matching competitive WAMs without additional policy pretraining across different simulator and real-world experiments. It also reduces FLOPs to 1/6 and latency to 1/4 of video-based WAMs. Attention analysis further shows that editing caches focus on task-relevant change regions, supporting image editing as an effective alternative to video-based world-action modeling.
Original Article
View Cached Full Text

Cached at: 06/20/26, 02:30 PM

Paper page - ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?

Source: https://huggingface.co/papers/2606.19531

Abstract

ImageWAM demonstrates that pretrained image editing models can effectively replace video generation in world action models for robot control, achieving better performance with reduced computational costs.

World Action Models(WAMs) commonly rely onvideo generationto bridgevisual world modelingandrobot control. However, video-based WAMs face three coupled limitations: dense multi-frame future tokens make inference costly, full video prediction spends capacity on action-irrelevant temporal and appearance details, and long-horizon future imagination may introduce errors that misleadaction prediction. These issues raise a simple question: Does world action model really needvideo generation? We propose ImageWAM, a simple WAM framework that repurposes pretrainedimage editingmodels for robotaction prediction. In contrast tovideo generation,image editingprovides a better-matched prior: it only needs to model a target-frame transformation, focuses on action-relevant current-to-target visual differences, and grounds task instructions to localized visual changes through edit pretraining. In practice, ImageWAM does not decode the target frame at inference time; instead, it conditions aflow-matching action experton theKV cachesproduced by image-editingdenoising, using them as a compact world-action context. ImageWAM outperforms standard VLA baselines and matching competitive WAMs without additional policy pretraining across different simulator and real-world experiments. It also reduces FLOPs to 1/6 and latency to 1/4 of video-based WAMs.Attention analysisfurther shows that editing caches focus on task-relevant change regions, supportingimage editingas an effective alternative to video-based world-action modeling.

View arXiv pageView PDFProject pageGitHub27Add to collection

Get this paper in your agent:

hf papers read 2606\.19531

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper3

#### yuyangalin/ImageWAM-FLUX.2-4B-RoboTwin Robotics• Updated1 day ago • 5 #### yuyangalin/ImageWAM-FLUX.2-4B-LIBERO Robotics• Updated1 day ago • 8 #### yuyangalin/ImageWAM-FLUX.2-9B-LIBERO Robotics• Updated1 day ago • 4

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2606.19531 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.19531 in a Space README.md to link it from this page.

Collections including this paper1

Similar Articles

World Action Models: The Next Frontier in Embodied AI

Hugging Face Daily Papers

This survey paper introduces World Action Models (WAMs), a unified framework for embodied AI that integrates predictive state modeling with action generation. It provides a taxonomy of existing methods, analyzes the data ecosystem, and outlines evaluation protocols for this emerging paradigm.

τ_0-WM: A Unified Video-Action World Model for Robotic Manipulation

Hugging Face Daily Papers

τ_0-WM is a unified video-action world model for robotic manipulation that integrates policy learning, video prediction, and action evaluation using a shared video diffusion backbone. It shows superior performance on challenging long-horizon and fine-grained tasks.