Thinker-WM: A Unified World Model with Shared Multimodal Representations for Embodied Intelligence
Thinker-WM: A Unified World Model with Shared Multimodal Representations for Embodied Intelligence
This research presents Thinker-WM, a unified world model that jointly models RGB, metric depth, and robot action through a shared intermediate representation. The model supports multiple functions including video-action generation, world prediction, and policy inference, and achieves top performance on several embodied intelligence benchmarks.
Thinker-WM introduces a unified world model that integrates visual, geometric, and action modalities through a shared dynamics transformer, demonstrating strong performance on benchmark tasks and showing the value of shared representations for embodied intelligence.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported duration
≈20,000 hours
View original evidence
duce a lightweight motion probe that shows the frozen visual representation contains recoverable motion information. Thinker-WM is trained on a heterogeneous corpus comprising approximately 20K hours of egocentric video, 10K hours of real-robot data, and 1K hours of simulation data, including approximately 4K hours of real-robot data with aligned depth supervision. Extensive experiments demonstrate strong performance
Open source S5Reported duration
10,000 hours
View original evidence
duce a lightweight motion probe that shows the frozen visual representation contains recoverable motion information. Thinker-WM is trained on a heterogeneous corpus comprising approximately 20K hours of egocentric video, 10K hours of real-robot data, and 1K hours of simulation data, including approximately 4K hours of real-robot data with aligned depth supervision. Extensive experiments demonstrate strong performance
Open source S5Reported duration
1,000 hours
View original evidence
duce a lightweight motion probe that shows the frozen visual representation contains recoverable motion information. Thinker-WM is trained on a heterogeneous corpus comprising approximately 20K hours of egocentric video, 10K hours of real-robot data, and 1K hours of simulation data, including approximately 4K hours of real-robot data with aligned depth supervision. Extensive experiments demonstrate strong performance
Open source S5Reported duration
≈4,000 hours
View original evidence
duce a lightweight motion probe that shows the frozen visual representation contains recoverable motion information. Thinker-WM is trained on a heterogeneous corpus comprising approximately 20K hours of egocentric video, 10K hours of real-robot data, and 1K hours of simulation data, including approximately 4K hours of real-robot data with aligned depth supervision. Extensive experiments demonstrate strong performance
Open source S5
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
duce a lightweight motion probe that shows the frozen visual representation contains recoverable motion information. Thinker-WM is trained on a heterogeneous corpus comprising approximately 20K hours of egocentric video, 10K hours of real-robot data, and 1K hours of simulation data, including approximately 4K hours of real-robot data with aligned depth supervision. Extensive experiments demonstrate strong performance
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org