From carrying out tasks to weighing values, Yunji equips robots with a "decision-making layer."

Recently, at the 2026 World Artificial Intelligence Conference (WAIC), Yunji unveiled the “World Value Model for Human-Machine Symbiosis,” and showcased products including its embodied intelligence single-arm collaborative robot and the UP200 industrial chassis.

According to Yunji’s definition, the World Value Model is a human-machine collaborative intelligent brain aimed at physical spaces. It doesn’t just control a single robot to complete a task, but coordinates people, robots, devices, and business systems in real-world scenarios, dynamically allocating resources based on task priorities, user experience, and operational returns.

This is a layer of “high-level decision system” Yunji is attempting to add to service robots.

Over the past decade or so, the service robot industry has mainly tackled whether robots can reliably operate in real-world environments. Positioning and navigation, obstacle avoidance, elevator usage, automatic charging, fault recovery, and multi-machine operation have determined if robots can work long-term in hotel corridors, hospital floors, and office buildings.

But as the number of robots increases, and the devices and tasks they connect to become more complex, addressing “whether a single machine can complete a task” is no longer enough.

Take hotels as an example—at the same time, there may be demands for water delivery, meal delivery, laundry delivery, and item collection. The number of robots is limited, and elevators, hallways, and laundry equipment might be in use. The system needs to know not only how robots get from guestrooms to the laundry room, but also which task is more urgent, which robot should be called, and whether staff intervention is needed if something goes wrong.

The World Value Model proposed by Yunji aims primarily to solve this layer of issues.

Conceptually, it’s not the same as the “World Model” often discussed in the industry.

The World Model typically focuses on understanding and predicting the physical world. Through learning the relationship between environmental state, actions, and outcomes, the model predicts how the world will change after an action is performed, and guides the robot accordingly.

Yunji’s World Value Model, however, does not aim to simulate the physical world as its primary goal; it is more akin to a value decision layer built atop the world model, robot control systems, and business systems.

“We’re optimizing across a longer task chain and a larger service scale,” said Duan Yanbiao, Vice President of Product at Yunji, to Wall Street Insights. “The world model uses predictive thinking, which can offer capabilities and guidance to the execution layer of robots. However, the focus of our model is ‘optimizing the entire task at a higher level.’”

In other words, the World Value Model seeks to answer “under current conditions, which task is more worthwhile to perform.”

To achieve this, Yunji divides the World Value Model into a four-layer architecture from T1 to T4, corresponding to scenario understanding, task planning, resource scheduling, and atomic execution.

T1 is responsible for understanding user needs and situational status; T2 splits a service goal into continuous task processes; T3 determines whether and when tasks are executed, and by whom; T4 invokes robots, mechanical arms, elevators, access control, and business systems to perform specific actions, feeding back results to the system.

Take hotel laundry as an example. When a guest requests laundry service, the system must first identify information about the items, pick-up/drop-off times, and special requirements, and obtain the real-time status of delivery robots, mechanical arms, washing machines, elevators, and passageways. The task is then broken down into several steps: pick-up, transport, handover, loading, washing, unloading, and return.

During execution, the system must also decide whether to handle the task immediately, which delivery robot and laundry device to call. If the elevator is congested, equipment fails, or a new urgent task arises, the system may need to reorder steps or assign certain parts to humans instead.

This contrasts with traditional robots that follow fixed rules to perform single delivery tasks. It handles not only action chains but also conflicts between tasks, real-time resource changes, and the division of labor between humans and machines.

This model is backed by operational data Yunji has accumulated in hotels and other venues over the years.

Yunji’s products have entered more than 40,000 hotels and around 200 hospitals worldwide, accumulating 1.1 billion service interaction records. Unlike internet models, this data isn’t just user clicks or textual conversations—it includes robot tasks, spatial status, device calls, exception handling, and human takeover in real operations.

But whether the World Value Model can truly be established depends on whether it can consistently make more economical, demand-driven decisions across different scenarios.

Risk Disclaimer and Exemption ClauseThe market has risks; investment should be approached with caution. This article does not constitute personal investment advice, nor does it take into account individual users’ specific investment goals, financial situation, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their particular circumstances. Investing accordingly is at your own risk.