InternW0-$Δ$: A World Action Model Bridging Predictive Dynamics and Actions with 20K+ Hours of Open Data

2026-09-25

AI Summary

This study addresses the challenge of integrating multimodal priors for general-purpose robotic manipulation by proposing a Unified World-Action Model. Methodologically, we introduce a Causal Imprint mechanism to achieve predictive representations without future rollout, and adopt a hybrid Transformer architecture that incorporates frozen vision-language model (VLM) semantic guidance, 4D foundation model distillation, and multi-source data alignment techniques. Furthermore, we construct the largest heterogeneous open-source corpus to date, comprising 20,000 hours of data. Experimental results demonstrate that the proposed model outperforms existing methods across both simulated and real-world platforms. To facilitate further research, we comprehensively release the source code, model weights, and data processing pipelines.

Abstract

World Action Models (WAMs) jointly model visual dynamics and action generation for generalist robot manipulation. A central challenge is to integrate priors from large-scale pretrained models---including visual dynamics, scene semantics, geometry, and motion---into a unified framework for robot action generation. We introduce InternW0-$Δ$, a unified WAM pretrained on a heterogeneous corpus that outperforms prior methods across simulation benchmarks and real-robot platforms. InternW0-$Δ$ combines pretrained visual dynamics, scene-level semantics, 4D geometric and motion priors, and action generation within a Mixture-of-Transformers (MoT) framework. A pretrained video expert and an action expert interact under semantic guidance from a frozen VLM, while a pretrained 4D foundation model injects geometric and motion priors through training-only distillation. We further introduce Causal Imprint, which learns future-relevant scene changes from training-only future supervision and provides predictive representations directly to the action expert without future-video rollout at inference. For large-scale joint training, we construct a heterogeneous corpus of robot demonstrations, UMI data, egocentric human demonstrations, and Ego2Robot data, curated and aligned under a common state-action representation. The resulting corpus contains over 20K hours of processed training data, to our knowledge the largest open-source corpus of its kind. We pretrain InternW0-$Δ$ on this corpus and demonstrate strong performance across simulation benchmarks and real-robot platforms. We will open source the training code, model weights, infrastructure, data-processing pipeline, and processed data where licenses permit. Project page: https://internrobotics.github.io/InternW0-Delta/

Problem

• World Action Models
• robot manipulation
• visual dynamics
• action generation
• pretrained priors

Innovation

• World Action Model
• Mixture-of-Transformers
• Causal Imprint
• Knowledge Distillation
• Heterogeneous Corpus