2025 STCSM Key Technology R&D Plan "New Generation Information Technology" Program

2025 STCSM Key Technology R&D Plan "New Generation Information Technology" Program

上海市科学技术委员会2025年度关键技术研发计划“新一代信息技术”项目

September 9, 2025

The Official Announcement in Chinese

Sponsor:
Science and Technology Commission of Shanghai Municipality (STCSM)

About: 
To accelerate the development of a globally influential science and technology innovation hub and strengthen the city's role as a source of scientific and technological innovation in the field of next-generation information technology, the STCSM hereby issues this guideline.

Funding Areas:

Area One: Frontier Technology Research on AI-Driven Paradigm Shifts in Scientific Research

• Topic 1: Common Support Technologies for Scientific Computing
  • Direction 1: Frontier Technology Research on Fundamental Operators in Scientific Computing
    Research Objectives: To address the need for indigenous development of fundamental operator libraries for scientific computing, developing foundational operator libraries based on domestically produced GPUs, covering linear algebra, numerical computation, and optimization solvers. The key performance metrics of these operators meet or exceed those of leading international commercial libraries. Conducting application validation in real-world scientific computing scenarios and providing scientific computing solutions to organizations in at least two distinct fields.
    Funding Amount: Up to RMB 2,000,000/project. Up to 1 project will be funded.
    Project Period: December 1, 2025 to November 30, 2027.
     
  • Direction 2: Frontier Research on Large-Model Inference Acceleration for Scientific Computing
    Research Objectives: Addressing the challenge that existing large-model inference acceleration methods struggle to balance efficiency and accuracy in complex scientific computing, this research develops a fine-grained inference acceleration mechanism for large models tailored to scientific computing. Based on domestic computing platforms, it achieves inference acceleration ratios no less than those of international mainstream algorithm libraries while maintaining equivalent model accuracy. The research outcomes form an algorithm library open-sourced to the industry and undergo application validation in at least two real-world inference scenarios.
    Funding Amount: Up to RMB 2,000,000/project. Up to 1 project will be funded.
    Project Period: December 1, 2025 to November 30, 2027.
     
• Topic 2: Data Governance and Model Agents
  • Direction 2: Development of Scientific Data Governance Toolchain and Research on New Model Architecture
    Research Objectives: Addressing challenges such as the scarcity of high-quality scientific data and the difficulty of multimodal, cross-scale characterization, this research focuses on developing efficient AI-driven governance technologies for scientific experimental data. It constructs an AI-enabled data governance toolchain to build high-quality scientific datasets in fields including biology, medicine, and chemistry. Additionally, it investigates innovative algorithms and architectures tailored to the disciplinary characteristics of corresponding vertical domains. The dataset features at least 100,000 effectively annotated entries, with key evaluation metrics of the new model achieving internationally leading levels.
    Funding Amount: Up to RMB 7,000,000/project. Up to 3 projects will be funded. 
    Project Period: December 1, 2025 to November 30, 2027.
    Team Requirements: Implement a dual-responsibility system involving both domain scientists and AI scientists. When submitting the application, clearly define the division of responsibilities between the two responsible parties in the project feasibility plan.
     
  • Direction 3: Frontier Research on Artificial Intelligence Agents
    Research Objectives: To address challenges such as low efficiency in manual literature reading comprehension and difficulties in proposing innovative hypotheses during scientific research, developing an intelligent agent capable of understanding scientific literature and generating high-quality hypotheses. This agent has undergone application validation in fields including biology, medicine, and chemistry, achieving key metrics at internationally leading levels.
    Funding Amount: Up to RMB 3,000,000/project. Up to 1 project will be funded. 
    Project Period: December 1, 2025 to November 30, 2027.
     
  • Direction 4: Frontier Research and System Validation of Intelligent Agents for Scientific Experiments
    Research Objectives: To address challenges such as high reliance on experience and low efficiency in scientific experiments, researching and developing an intelligent agent capable of automatically designing experimental workflows, efficiently analyzing experimental data, and intelligently orchestrating experimental protocols. This agent has undergone application validation in fields including biology, medicine, and chemistry, achieving key metrics at internationally leading levels.
    Funding Amount: Up to RMB 3,000,000/project. Up to 1 project will be funded. 
    Project Period: December 1, 2025 to November 30, 2027.
     
• Topic 3: Research on Next-Generation Frontier Technologies
  • Direction 1: Science-inspired AI Free Exploration
    Research Objectives: Addressing current challenges in artificial intelligence models—including robustness, computational efficiency, and explainability—researchers are developing science-inspired disruptive algorithms, models, and tools to explore new pathways for AI transformation.
    Funding Amount: Up to RMB 500,000/project. Up to10 projects will be funded. 
    Project Period: December 1, 2025 to November 30, 2026.
    Requirements for PIs: The PI must not be older than 35 years old (born after January 1, 1990). In principle, the age requirements for other core researchers are the same as above. Promising teams will receive phased rolling support through competitive selection, targeted commissioning, and other methods.

Area Two: Frontier Research and Applications in Embodied Intelligence

• Topic 4: Key Technologies for Embodied Intelligent Body Modules
  • Direction 1: Multimodal Tactile Perception and Identification Technology and Module Development
    Research Objectives: For embodied intelligent manipulation tasks involving flexible objects, research state perception and parameter identification techniques for multi-stiffness flexible bodies. Develop a vision-tactile multimodal perception module with a vision-tactile contact force perception error ≤0.05N and flexible unit stiffness identification accuracy ≥90%. Complete embodied intelligent manipulation tasks for five or more types of flexible objects.
    Funding Amount: Up to RMB 3,000,000/project. Up to 1 project will be funded.
    Project Period: December 1, 2025 to November 30, 2027.
  • Direction 2: Development of Terminal-side Embodied Intelligent Model Operation Module
    Research Objectives: Aiming to meet the application needs of embodied intelligence models on robotics, developing a lightweight embodied intelligence model and its operational module that supports multimodal perception and interaction strategy generation. The model boasts over one billion parameters, with key core components being fully domestically produced. The model must achieve a success rate of at least 90% for missions it has already seen.
    Funding Amount: Up to RMB 3,000,000/project. Up to 1 project will be funded.
    Project Period: December 1, 2025 to November 30, 2027.
     
• Topic 5: Embodied Intelligent Operating System and Toolchain Technology
  • Direction 1: Research on Embodied Intelligent Body End-side Model Middleware Technology
    Research Objectives: To address the challenges of poor generalization and low operational efficiency in edge-side intelligent agents, developing an operating system middleware that supports long-term task decomposition and skill model orchestration, as well as concurrent scheduling and high-performance inference of multiple models on edge-side XPUs. This enables intelligent orchestration of at least three minute-level tasks in scenarios such as home services and manufacturing, achieving 50Hz model action output on domestic computing power.
    Funding Amount: Up to RMB 3,000,000/project. Up to 1 project will be funded.
    Project Period: December 1, 2025 to November 30, 2027.
     
  • Direction 2: Development of Embodied Operation-Based Data Acquisition Technology and System
    Research Objectives: To address the demand for efficient collection of authentic and reliable data, focusing on the research, development, and application of embodied operation data acquisition systems. To achieve breakthroughs in key technologies such as time synchronization and spatial alignment for multi-modal data acquisition encompassing position, force, touch, and vision. The system meets performance metrics including ≥50 degrees of freedom for motion capture, ≤0.5N accuracy for force/tactile sensing, and ≤0.5kg total weight. Its time synchronization and spatial alignment capabilities must reach internationally leading standards.
    Funding Amount: Up to RMB 3,000,000/project. Up to 1 project will be funded.
    Project Period: December 1, 2025 to November 30, 2027.
     
  • Direction 3: Development of Embodied Intelligence Simulation and Data Synthesis Engine
    Research Objectives: To address the high costs and challenges associated with collecting real-world embodied intelligence data, developing an embodied intelligence simulation and data synthesis engine. This engine supports the construction of highly realistic scenarios and the automated synthesis of multimodal embodied intelligence data. It enables GPU-based parallel computing, achieving simulation speeds of up to 10 million frames per second (FPS), and supports the training and application of embodied intelligence models for over 10 types of fine-grained operational tasks.
    Funding Amount: Up to RMB 3,000,000/project. Up to 1 project will be funded.
    Project Period: December 1, 2025 to November 30, 2027.

Please refer to the official announcement for more details about each research areas, topics and directions. 

Eligibility:

• Applicants who have served as PIs on two or more ongoing science and technology projects funded by the STCSM are ineligible to apply as PIs;
• For applicants proposing new projects based on previous projects funded by municipal financial resources or other organizations (e.g., the Ministry of Science and Technology, the National Natural Science Foundation of China, etc.), the similarities and differences between the two, and the relationship between inheritance and development should be clearly stated.

Please refer to the official announcement for more details about eligibility requirements.

Important Dates: