Jiangsu Issues Action Plan for Building High-Quality Datasets for Physical AI and World Models (2026–2028)

In October 2026, the Jiangsu Provincial Data Bureau, the Provincial Development and Reform Commission, the Provincial Department of Science and Technology, and the Provincial Department of Industry and Information Technology recently formally issued the "Jiangsu Province Physical AI and World Models High-Quality Dataset Construction Action Plan (2026–2028)."

Jiangsu Province Physical AI and World Models High-Quality Dataset Construction Action Plan (2026–2028)

This Action Plan is based on Jiangsu Province's foundation as a major manufacturing province and its advantages in science, education, and talent. It follows the trend of artificial intelligence evolving from digital simulation to physical interaction and from digital intelligence to spatial intelligence. It focuses on key directions such as building real-machine interaction datasets, expanding simulation and synthetic data, and tackling cutting-edge scientific research datasets, and accelerates the construction of a virtuous, mutually reinforcing pattern in which "scenarios pull data, data drives models, models empower applications, and applications create value."

The Action Plan sets out the overall development goals: by the end of 2028, centering on the "1650" industrial system, the "51010" strategic emerging industry clusters, and the AI application needs of the "6+6" industries, Jiangsu will build high-quality physical AI datasets covering more than 100 key scenarios, and will build more than 10 high-quality datasets oriented toward the frontier direction of world models that possess causal consistency, spatiotemporal continuity, and application verifiability. These will support models in improving physical cognition, real-time deduction, and iterative optimization capabilities; achieve breakthroughs in key technologies such as multi-source heterogeneous semantic alignment and fusion processing, ontology modeling, and multimodal data synthesis and simulation enhancement; form a batch of standards for data collection, annotation, and evaluation; and create a batch of typical application scenarios in which data drives the innovative development of physical AI.

The action specifies a list of 5 major items and 14 key tasks, and breaks down responsibilities to departments to ensure that the construction of high-quality datasets has goals, concrete measures, and can be implemented.

Sort out the data needs of physical AI and world models across the province for scenario sensing, action interaction, and physical simulation; incorporate the real tasks that need to be solved at industrial sites, model prediction deviations, abnormal working conditions, and actual execution feedback into the sources of demand; rely on the Jiangsu Province High-Quality Dataset Management Service Platform to dynamically update the demand list, connect to the National High-Quality Dataset Management Service Platform and publish it online, and guide real-scene collection and simulation generation to precisely match demand. (Led by the Cyberspace Administration of the Provincial Party Committee and the Provincial Data Bureau; relevant departments of each designated city shall carry out work according to their responsibilities. The following key tasks all involve each designated city and will not be listed repeatedly.)

2. Promote in-depth collaboration between model R&D institutions and industry enterprises.

Promote collaboration between R&D institutions of physical AI and world models in fields such as intelligent manufacturing, embodied intelligence, and autonomous driving and industry enterprises; focus on the data bottlenecks constraining the iteration of physical AI and world models; build virtual-real training and validation environments that support algorithms such as reinforcement learning; jointly formulate plans for data collection, annotation, simulation enhancement, and virtual-real integration; rely on interaction feedback from perception, decision-making, and control links in real scenarios to support analysis of the causes of prediction failures and targeted supplementary data accordingly; and build a closed-loop iteration system between data and models. (Led by the Provincial Data Bureau, with relevant departments responsible according to their division of duties.)

3. Establish a cooperation mechanism between data annotation enterprises and model R&D institutions.

Encourage R&D institutions of physical AI and world models to open demand interfaces, and attract multiple parties to participate in the construction of high-quality datasets through methods such as OPC crowdsourcing, holding high-quality dataset developer innovation competitions, forming industry high-quality dataset innovation consortia, promoting public data authorized operation entities to collaborate with annotation enterprises, and deepening industry-education integration and co-construction, thereby forming a collaborative model of "models propose demand—society builds data—market verifies results." (Led by the Provincial Data Bureau, with relevant departments responsible according to their division of duties.)

(II) Tackle the supply and application of datasets in key fields

4. Construction of real-machine interaction datasets.

(1) Consolidate the foundation of high-quality datasets for traditional advantageous industries. Focus on advantageous fields such as high-end equipment, new energy, new power equipment, high-tech ships and marine engineering equipment, and biomedicine; support leading enterprises in the industry, key industrial internet platform enterprises, and key industrial software enterprises to focus on real-machine interaction scenarios in production and manufacturing; integrate real-time data such as production processes and operating environments; build industry high-quality datasets oriented toward the technical directions of physical AI and world models; and empower the digital and intelligent transformation and upgrading of traditional advantageous industries. (The Provincial Department of Industry and Information Technology and the Provincial Data Bureau are responsible according to their division of duties.)

(2) Cultivate high-quality datasets for emerging pillar industries. Around fields such as integrated circuits, the low-altitude economy, new energy storage, and intelligent robots, advance the coordinated application of physical mechanism modeling, defect sample expansion, and real-machine data; form datasets covering the entire process of R&D, manufacturing, testing, operation, and maintenance; empower scenarios such as chip design, low-altitude intelligent networking and scheduling, coordinated optimization of source-grid-load-storage, and generalized robot operation; support the deep integration of emerging pillar industries with physical AI; and accelerate the upgrading of industrial capacity. (The Provincial Department of Industry and Information Technology, the Provincial Development and Reform Commission, the Provincial Department of Transportation, the Provincial Department of Science and Technology, and the Provincial Data Bureau are responsible according to their division of duties.)

(3) Prospectively lay out high-quality datasets for future industries. Around fields such as quantum technology, biomanufacturing, and sixth-generation mobile communications, rely on high-level platforms such as the Yangtze River Delta Industrial Innovation Center for Quantum Technology and the Zijinshan Laboratory to increase efforts in experimental data collection, governance, and integration; build high-quality datasets; empower model training in quantum physical modeling, biomanufacturing control, and integrated sensing, communication, and computing; and support frontier exploration and key technology validation in future industries. (The Provincial Department of Science and Technology, the Provincial Development and Reform Commission, the Provincial Health Commission, the Provincial Medical Products Administration, the Provincial Communications Administration, and the Provincial Data Bureau are responsible according to their division of duties.)

5. Expansion of simulation and synthetic data.

Focusing on scenarios such as industrial sites, high-risk operations, logistics and warehousing, medical surgery, emergency rescue, and multi-hazard coupling, advance multi-physics simulation, real-time sensing, contact operations, domain randomization, Real2Sim (conversion of real-world data into simulation environments) transfer, and joint training; build high-fidelity digital twin simulation environments; form high-quality synthetic datasets; and support closed-loop iteration of related physical AI products. (The Provincial Department of Industry and Information Technology, the Provincial Department of Transportation, the Provincial Department of Natural Resources, the Provincial Department of Housing and Urban-Rural Development, the Provincial Department of Emergency Management, the Provincial Meteorological Bureau, the Provincial Department of Science and Technology, and the Provincial Data Bureau are responsible according to their division of duties.)

6. Breakthroughs in frontier scientific research datasets.

Grounded in the frontier development paradigm of AI for Science (artificial intelligence empowering scientific research) and oriented to the needs of world model training, validation, and evaluation, focus on spatiotemporal alignment and integrated annotation of multi-source scientific data; promote the construction of a data supply and application system covering the entire chain of basic research, technical breakthroughs, pilot validation, and industrial application; and support world models in learning, validating, and iteratively optimizing physical laws, spatiotemporal relationships, and causal mechanisms. (The Provincial Department of Science and Technology, the Provincial Department of Education, and the Provincial Data Bureau are responsible according to their division of duties.)

(III) Deepen enterprise pilot cultivation through point-to-area approaches

7. Solidly advance the early and pilot implementation of data annotation and high-quality dataset construction.

Accelerate the construction of Nanjing's national early and pilot implementation of data annotation (domain innovation—industrial manufacturing direction), support the construction of Nanjing's data annotation industry and high-quality dataset public service platform, support Wuxi, Suzhou, and Suqian in building provincial-level annotation bases in the fields of industrial manufacturing, smart transportation, and embodied intelligence, and encourage regions with suitable conditions to develop the data annotation industry according to local conditions, forming a differentiated development layout. (The Provincial Data Bureau and the Provincial Department of Industry and Information Technology take the lead, and relevant departments are responsible according to their division of duties.)

8. Deepen the demonstration effect of dataset construction by provincial state-owned enterprises.

Give play to the demonstration and leading role of provincial state-owned enterprises, improve the data classification and grading management mechanism, build a data resource system covering the entire business chain, promote key enterprises to mine scenario data in fields such as energy and road networks, build high-quality datasets for physical AI, drive industrial chain collaboration through consortia, and establish open sharing and inclusive mechanisms to provide authorized use services to small and medium-sized enterprises, universities, and startup teams in accordance with regulations, reduce the cost of data acquisition and use, and promote the reuse of data across the entire industry. (The Provincial Data Bureau and the Provincial State-owned Assets Supervision and Administration Commission are responsible according to their division of duties.)

9. Establish a dynamic collection and tiered cultivation mechanism for datasets.

Build a special database for dataset projects, routinely solicit high-quality dataset construction projects in physical AI and world models within the province for inclusion in the database, and implement tiered cultivation management. For reserve projects, strengthen demand matching and resource introduction and accelerate project incubation. For projects under construction, strengthen quality assessment and technical support to promote quality and efficiency improvement. Organize provincial-level solicitation and selection of high-quality dataset construction projects, carry out application validation and promotion and replication for mature projects, and form a demonstration-driven effect. (The Provincial Data Bureau takes the lead, and relevant departments are responsible according to their division of duties.)

(IV) Advance technical breakthroughs and infrastructure construction

10. Carry out key technical breakthroughs in physical AI and world model data.

According to local conditions, systematically identify key technical bottlenecks in the collection, processing, circulation, and integrated application of physical AI and world model data, and focus on breakthroughs in technologies such as multi-source heterogeneous semantic alignment and fusion processing, ontology modeling, and multimodal data synthesis and simulation enhancement. Oriented to the frontier direction of world models, advance technical breakthroughs in multimodal spatiotemporal alignment, cross-domain adaptive transfer, and causal consistency modeling. Explore the establishment of a large model evaluation and annotation technical system for physical AI and world models. (The Provincial Data Bureau takes the lead, and the Provincial Department of Science and Technology is responsible according to its division of duties.)

11. Build physical AI data production and training bases.

Lay out physical AI data production and training bases, build simulation validation environments, provide capabilities such as physical scenario data collection, specialized annotation, extreme scenario synthesis, and virtual-real integration validation, and support world model training, testing, and iterative optimization. (The Provincial Data Bureau and the Provincial Development and Reform Commission take the lead, and relevant departments are responsible according to their division of duties.)

(V) Improve the circulation, transaction, and standards and evaluation systems

12. Improve the data circulation and transaction system.

Clarify subject responsibilities, rights and obligations, and codes of conduct. Relying on data circulation service institutions such as the provincial data exchange, improve service functions such as registration, evaluation, matching, delivery, and settlement. Explore a value measurement system based on tokens. Support the construction of industry-specific circulation platforms and promote efficient circulation of data in vertical fields. Promote the construction of a trusted circulation environment and strengthen the application of technologies such as privacy computing and blockchain. (The Provincial Data Bureau takes the lead, and relevant departments are responsible according to their division of duties.)

13. Improve the standards and specifications system.

Formulate guidelines for high-quality datasets for physical AI, build a standards framework for data annotation and high-quality datasets, strengthen the construction of sub-systems of standards in areas such as spatiotemporal alignment, multimodal data governance, entity interaction annotation, and simulation data tool platforms, organize the implementation of national standards, and encourage reliance on evaluation and certification institutions to build model capability evaluation systems for industries and special scenarios. (The Provincial Data Bureau takes the lead, and relevant departments are responsible according to their division of duties.)

14. Strengthen data security and ethical compliance review.

Strictly implement data classification and graded protection requirements. For data such as medical clinical, genetic, and biomanufacturing data, carry out ethical review in accordance with the law and implement informed consent requirements. For data such as geographic surveying and mapping and urban video, strictly implement confidentiality provisions. Formulate security compliance guidelines, clarify responsibilities across the entire process of collection, storage, use, sharing, and destruction, and have industry regulatory authorities implement regulatory responsibilities according to their duties. (The Provincial Data Bureau takes the lead, and relevant departments are responsible according to their division of duties.) List of Key Projects and List of Cultivation Projects for High-Quality Dataset Construction in Jiangsu Province in 2026 (Physical AI and World Models Direction)

The List of Key Projects and the List of Cultivation Projects for High-Quality Dataset Construction in Jiangsu Province in 2026 (Physical AI and World Models Direction) has recently been officially issued.

High-quality datasets are an important foundation for artificial intelligence to move from digital space to the physical world. At present, physical AI and world models are becoming a new focus of global artificial intelligence competition, creating an urgent demand for high-quality datasets. To this end, Jiangsu has targeted the physical AI and world models direction and issued the 2026 provincial list of key projects and list of cultivation projects for high-quality dataset construction, promoting the large-scale construction and open sharing of high-quality datasets, effectively transforming data advantages into industrial advantages, and taking the lead and setting an example in the development of physical AI and world models.

In the 2026 Provincial List of Key Projects for High-Quality Dataset Construction (Physical AI and World Models Direction), a total of 39 projects were selected, including the high-quality dataset construction project for embodied intelligent welding robots.

In the 2026 Provincial List of Cultivation Projects for High-Quality Dataset Construction (Physical AI and World Models Direction), a total of 79 projects from 12 prefecture-level cities were selected. The projects cover 15 tracks in total: modern agriculture, industrial manufacturing, urban governance, artificial intelligence, embodied intelligence, intelligent driving, healthcare, meteorological services, cultural tourism, housing construction, low-altitude economy, public safety, green and low-carbon, natural resources, and emergency management.

Terminology

Physical AI: Refers to intelligent systems that integrate perception, cognition, and control capabilities, enabling agents to possess organism-like perception, reasoning, decision-making, and action capabilities, to understand physical laws such as gravity, friction, and collision as well as causal and spatiotemporal relationships, and to be able to trial and error and learn under uncertain conditions, and to bear responsibility for the consequences of their physical actions.

World Model: Refers to a model that, based on the physical laws of the real world, reconstructs the environment within the system and simulates its evolution process, predicts the possible outcomes of different actions, and supports action planning, counterfactual reasoning, and closed-loop correction.

www.smartcity.team

Disclaimer: Please indicate the source as "Smart City Industry Analysis (www.smartcity.team)" and the relevant author when reprinting. Except for original content, some information on this site comes from the Internet. If the relevant content involves infringement of rights, please contact the contact information at the bottom to delete it.

[Digital Technology Special Research] Community

In-depth peer exchange, focused on research and material accumulation in the smart city and digital technology industries.

Continuously updating the latest industry white papers, industry research, policy developments, and implementation solutions to help industry practitioners efficiently accumulate knowledge and share experience.

Massive premium content continues to iterate, providing cutting-edge perspectives, connecting industry talent, carrying out case co-creation and resource exchange, and jointly building a high-quality industry exchange atmosphere.

#post-##

Recommended for you

Guizhou "Model-Data Factory" Construction and Operation Work Plan (2026-2028): Coordinate computing power, data, applications, and industry; by 2028, strive to fully build a benchmark park, with industrial space exceeding 40,000 square meters, gathering more than 100 enterprises, and park revenue exceeding 1 billion yuan

Wuhan Special Construction Fund Management Measures for High-Quality Datasets: Support a batch of industry high-quality dataset construction projects each year, granting rewards of up to 2 million yuan based on 25% of the project's actual R&D investment

"Several Measures of Shanghai to Promote High-Level Opening-Up and Facilitate Cross-Border Data Flow": Jointly issued by the Cyberspace Administration and the Data Bureau, 10 measures to facilitate cross-border data flow

China Meteorological Administration "Meteorological Data Management Measures": Effective from November 1, 2026

Full text of the "Shandong Province Data Regulations": Effective from December 1, 2026, with 8 chapters and 66 articles, with key content covering coordinated data resource management, strengthening data rights protection, promoting data circulation and transactions, promoting industrial development and application, and strengthening data security and safeguards

Wenzhou Artificial Intelligence Training Data Trade Secret Protection Guidelines (2026 Edition): Divided into three confidentiality levels: core, key, and general. The data holder is the primary responsible entity for trade secret protection and needs to establish four lists: confidential data, confidential personnel, confidential carriers, and confidential areas