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

arXiv 2026 56.8 method, application

TLDR

ImageWAM repurposes image editing models for robot action prediction, outperforming video-based world action models with lower cost and latency.

Reasoning

The paper presents a novel approach that replaces video generation with image editing for world action models, achieving superior performance and efficiency in both simulator and real-world experiments. However, it lacks explicit discussion of limitations and potential dependencies on pretrained models.

Read-first score

Read-first score 56.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 36.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Citation impact 18%

94.6

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.94589443

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=analysis,baseline,experiment

Topical relevance 29%

51.4

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FoundationFrontierBridgeMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 423.

Keyword Scores

world model

generative world model

video world model

interactive world model

world dynamics prediction

model-based reinforcement learning world model

world simulator

Deep Analysis

Innovations

Proposes ImageWAM, a framework that repurposes pretrained image editing models for robot action prediction instead of video generation.
Uses KV caches from image-editing denoising as a compact world-action context, without decoding the target frame at inference.
Demonstrates that image editing provides a better-matched prior for world-action modeling: target-frame transformation, action-relevant visual differences, and localized visual changes through edit pretraining.

Methodology

ImageWAM repurposes a pretrained image editing model to produce KV caches during denoising, which condition a flow-matching action expert. It does not decode the target frame at inference time. The model is evaluated on simulator and real-world experiments, compared to standard VLA baselines and competitive WAMs.

Key Results

ImageWAM outperforms standard VLA baselines and matching competitive WAMs without additional policy pretraining across simulator and real-world experiments. It reduces FLOPs to 1/6 and latency to 1/4 of video-based WAMs.

Resources

Paper