On June 15, Beijing ShengShu Technology Co., Ltd. (hereinafter referred to as “ShengShu Technology”) and Guanglun Intelligence (Beijing) Technology Co., Ltd. (hereinafter referred to as “Guanglun Intelligence”) announced a strategic partnership. The two parties will collaborate deeply on co-building data standards for general world models, co-building a high-quality world model evaluation system, and accelerating real-world deployment of world models, accelerating the path of general world models toward being “physically trustworthy, task-ready, and verifiable in real scenarios.”
Luo Yihang, CEO of ShengShu Technology, and Yang Haibo, Co-founder and President of Guanglun Intelligence, witnessed the signing
Over the past two years, multimodal large models have continued to break through in visual quality, generation duration, and motion consistency, transitioning from “content generation” toward “world models” — becoming the core foundation for understanding world states, predicting future evolutions, and planning real-world actions. As world models move toward the physical world, embodied intelligence is a key landing point. Robots executing tasks in real, dynamic, open scenarios must bridge the gap between perception and execution. The core value of world models no longer lies solely in generating content, but in whether they can understand physical laws, complete complex task reasoning, and provide support for real-world actions through the interaction between digital and physical spaces.
This collaboration is a powerful alliance at this industrial inflection point: ShengShu Technology leverages its deep technical expertise in general world models, with its World Action Model Motubrain at the core, building a general-purpose “brain” for embodied intelligence that unifies “predicting the world” and “driving action” into a single model; Guanglun Intelligence builds data and evaluation infrastructure for Physical AI around human data, simulation evaluation, and deployment feedback, leveraging its pioneering “Solve – Measure – Generate” integrated full-stack proprietary simulation platform to drive continuous iteration between real-world experience and digital world capabilities. The two parties will jointly advance three key directions: a high-quality world model data system, a reproducible world model evaluation system, and a real-world scenario validation and continuous iteration mechanism.
Data Foundation: Making World Models Better Understand the Physical World
The capability boundary of world models is fundamentally determined by the physical coverage and task diversity of their data. To meet the needs of general world model training and iteration, the two parties will combine ShengShu Technology’s technical expertise and heterogeneous multimodal data absorption capabilities in world models with Guanglun Intelligence’s accumulated data definition, quality systems, and best practices in human video data and synthetic simulation data. Together, they will explore data standards and construction methods for world models, drive efficient collaboration across heterogeneous data sources, and continuously improve world models’ ability to model physical space-time, causal logic, and dynamic evolution processes.
Co-building a Reproducible, High-Quality World Model Evaluation System
As world models enter the physical world, progress cannot be judged by subjective perception alone. Whether a model truly understands physical laws, completes concrete tasks, predicts action outcomes, and meets the conditions for deployment in real scenarios — all require a unified, stable, and reproducible evaluation system for validation. The two parties will leverage Guanglun Intelligence’s industrial-grade simulation evaluation platform RoboFinals to jointly explore evaluation tasks, metrics, and data quality standards for world models and embodied intelligence models, building a continuous iteration closed loop from capability diagnosis through data augmentation to model optimization. Evaluation results will serve not only as a scoring tool for model capabilities, but also as a critical basis for data selection, model optimization, and scenario validation.
Bringing World Models into Industrial Settings, Advancing Embodied Intelligence Toward Real Applications
The ultimate value of world models must be validated in real tasks and industrial scenarios. The two parties will conduct joint exploration around embodied intelligence and Physical AI application scenarios, covering key stages including data collection, task SOP conversion, simulation evaluation, real-machine verification, feedback assessment, and continuous iteration.
In this process, ShengShu Technology will leverage its general world model’s core capabilities in spatial understanding, physical law modeling, and end-to-end action decision-making, while Guanglun Intelligence will provide simulation scenarios, evaluation platforms, real-machine verification, and scenario delivery capabilities. Together, they will explore the transformation path from model capability to scenario application, driving validation, feedback, and iteration of world models and embodied intelligence in real industrial settings.
The strategic partnership between ShengShu Technology and Guanglun Intelligence is not merely a complementary alliance between a world model company and a Physical AI infrastructure company — it represents a key direction toward Physical AI. The two parties will continue to explore how world models can learn, reason, and validate in real-world environments, truly moving from the digital world to the physical world, accelerating into real industrial scenarios, and unlocking greater technical value and application potential.
Looking ahead, ShengShu Technology will build a unified model architecture to complete the full chain from perception to action, creating a closed loop that spans “understanding the world – predicting the world – acting in the world,” making general world models truly serve as the bridge connecting the digital and physical worlds.
About ShengShu Technology
Founded in 2023, ShengShu Technology is the world’s first company to achieve unification of the digital and physical worlds through general world models. The company is dedicated to building a universal intelligence framework for modeling, reasoning, prediction, and action based on the laws governing both digital and physical worlds. Adhering to first principles, ShengShu Technology trains foundational world models using visual and auditory information that naturally carries physical world representations, recreating the process of human cognition, simulation, and interaction with the physical world — ultimately achieving Artificial General Intelligence (AGI) that connects the digital and physical worlds.
About Guanglun Intelligence
Guanglun Intelligence is dedicated to building data and evaluation infrastructure for Physical AI. Leveraging its pioneering “Solve – Measure – Generate” integrated full-stack proprietary simulation platform, it delivers high-quality, scalable synthetic simulation data, industrial-grade simulation evaluation, human video data, and deployment feedback capabilities. As the world’s first embodied data unicorn, all top five global world model teams collaborate with Guanglun, achieving number one global delivery in synthetic simulation data, simulation evaluation, and human video data.