InternW0: A Foundational Physical World Model for Efficient Real-World Interactions

arXiv 2026 44.7 method, application

TLDR

InternW0 is a foundational physical world model combining video prediction and robot control via asymmetric flow-matching architecture, trained on heterogeneous data, and validated on real-world scientific tasks.

Reasoning

The paper presents a strong integration of visual dynamics and action at different timescales, backed by large-scale heterogeneous training data and real-world evaluations such as MOF synthesis and pipetting. However, the abstract is truncated and lacks quantitative baselines or detailed comparisons, making it difficult to assess novelty and effectiveness fully.

Read-first score

Read-first score 44.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 55.

Recency 5%

100

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

Methodology quality 15%

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

Topical relevance 25%

78.6

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 15%

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

Graph prestige 15%

5.7

PageRank-style prestige over the local citation graph built from available OpenAlex references. local_pagerank=0.0016

Citation impact 15%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 10%

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 687.

Keyword Scores

world model

interactive world model

video world model

world dynamics prediction

generative world model

world simulator

model-based reinforcement learning world model

Deep Analysis

Innovations

First instantiation of the InternW physical world model series, built around omnimodal interfaces, asynchronous multi-frequency processing, and local physical modeling under partial observations and external influences
Asymmetric video-action architecture that jointly learns future visual dynamics and continuous robot control using flow matching
High-capacity video expert for longer-horizon predictive context paired with a lightweight action expert operating at a faster timescale
Reuses layerwise K/V and adapts to newly observed states through observation-conditioned context routing instead of regenerating the future for every action update
Domain-specific interfaces and soft prompts supporting heterogeneous embodiments
Contact-aware post-training incorporating force and tactile signals for contact-rich manipulation
EgoLab dataset, a 275-hour real-laboratory egocentric dataset, included within approximately 7,200 hours of heterogeneous robot and egocentric training data

Methodology

InternW0 uses an asymmetric video-action architecture with flow matching to jointly learn future visual dynamics and continuous robot control. A high-capacity video expert supplies longer-horizon predictive context while a lightweight action expert operates at a faster timescale, reusing layerwise K/V with observation-conditioned context routing. The model is trained on about 7,200 hours of heterogeneous robot and egocentric data, including EgoLab, with contact-aware post-training using force and tactile signals.

Key Results

Evaluation spans simulation benchmarks and real-world scientific tasks, including a 15-stage metal-organic framework synthesis workflow and 5-stage contact- and force-aware dexterous manipulation for general-purpose quantitative pipetting.

Resources

Paper