InternW0-Δ: A World Action Model Bridging Predictive Dynamics and Actions with 20K+ Hours of Open Data
InternW0-Δ is a world action model pretrained on over 20,000 hours of open robot and human data.
InternW0-Δ is a world action model that jointly learns visual dynamics and robot action generation in a Mixture-of-Transformers framework. A video expert and an action expert use a frozen vision-language model for semantics, while a 4D foundation model distills geometric and motion priors during training. Causal Imprint gives the action expert future-relevant scene features without rolling out future video at inference. The authors report strong simulation and real-robot results and plan to open-source code, weights, infrastructure, and licensed processed data.
- Pretrained on more than 20,000 hours of robot, UMI, and egocentric data.
- Mixture-of-Transformers combines video, action, VLM semantics, and 4D priors.
- Causal Imprint supplies predictive features without inference-time video rollout.
- Code, weights, pipeline, and licensed processed data will be released.
- Organizations
- InternRobotics
- AI models
- InternW0-Δ
From the original article
1 min read at huggingface.co
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.
By Xingyu Miao, Zizun Li, Baole Fang · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.