2025 STCSM Special Program on Fundamental Research of “Large Models for Artificial General Intelligence" (First Round)

2025 STCSM Specialized Program on Fundamental Research of "Large Models for Artificial General Intelligence" (First Round)
2025年上海市“通用人工智能大模型”基础研究专项（第一批）

April 30, 2025

The Official Announcement in Chinese

Sponsor:

Science and Technology Commission of Shanghai Municipality (STCSM)

About: 　

In order to accelerate the construction of a science and technology innovation center with global influence and better play the role of the Shanghai Artificial Intelligence Laboratory as a science and technology innovation platform, the Shanghai Science and Technology Commission and the Shanghai Artificial Intelligence Laboratory jointly issued the 2025 Shanghai Specialized Program on Fundamental Research of "Large Models for Artificial General Intelligence" (first round).

Topics and Sub-areas:

Topic 1: Basic theories and technologies of artificial intelligence

Funding Amount: RMB 500,000/project. Each direction will support up to 2 projects.

Project Period: July 1, 2025, to June 30, 2026.

• Direction 1: Research on embodied autonomous learning algorithms
  • Research objectives: Aiming at the needs of autonomous evolution in embodied scenarios, the robot can autonomously explore the environment and self-learn new skills. It can complete deep-level decision-making and complex task execution through autonomous evolution and have generalization capabilities.
• Direction 2: Research on model autonomous evolution and multi-model co-evolution technology
  • Research objective: To address the problem of weak adaptability of large models in dynamic environments, realize the autonomous evolution of models and the collaborative evolution of multiple models in dynamic environments.
• Direction 3: Research on technology to enhance the reasoning capability of large models in scientific mission scenarios
  • Research objectives: To address the problems of insufficient reasoning capabilities and generalization of existing large models, break through the performance boundaries and scope of application of existing large model reasoning algorithms for scientific scenarios and conduct verification.
• Direction 4: Research on long-term planning technology for complex tasks of large-scale intelligent agents
  • Research objectives: To address the problems of insufficient planning ability and information loss in large-model intelligent agents when performing long-term tasks, improve the planning, decision-making and execution capabilities of intelligent agents in long-term complex tasks, and promote intelligent agents to achieve a higher level of autonomy.
• Direction 5: Research on traceable large model network architecture and algorithms
  • Research objectives: To address the problem of insufficient knowledge traceability and explainability of large language models, we will build an innovative network architecture and algorithm system that can be traced back to the source, and achieve accurate traceability of generated content.
• Direction 6: Research on algorithms to improve data intelligence density for large model training
  • Research objective: To address the problem of low data quality affecting the effectiveness of large model training, improve the intellectual density of low-quality data and verify its effectiveness in model training.
• Direction 7: Research on general operation algorithms driven by force-position hybrid decision control
  • Research objectives: To address the problem of insufficient precision and adaptability of pure position control in complex and ever-changing operating scenarios, a general force-position hybrid control strategy and its training algorithm that integrates multi-modal perceptions such as vision and touch are constructed to improve the robot's high-precision, robust and generalized execution capabilities for fine operating tasks.
• Direction 8: Research on general and efficient video understanding methods
  • Research objectives: To address the problems of high model complexity and insufficient deep understanding capabilities in long video understanding, and to develop a general, efficient and scalable basic model for video understanding.
• Direction 9: Research on multimodal generated content identification technology
  • Research objectives: To address the security risks brought by AI-generated content, and to improve the ability to stably identify multimodal generated content and the traceability of generation models.

Topic 2: Research and Evaluation of Innovation Management Mechanisms

Funding Amount: up to RMB 250,000/project. Each direction will support up to 3 projects.

Project Period: July 1, 2025, to June 30, 2026.

• Direction 1: Research on the design of innovative management mechanisms and implementation effectiveness evaluation methods for the "General Artificial Intelligence Large Model Basic Research Project"
  • Research objectives: To design innovative mechanisms in the implementation of this special project and produce high-quality effectiveness evaluation reports and innovative management mechanism recommendations.

Please refer to the official announcement for more details about each research direction/topic.

Application Eligibility:

• Each institution can only submit ONE application in each research objective under each research area.
• 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.

Additional notes for Awards:

• If a project is funded, the Shanghai Artificial Intelligence Laboratory will engage in consultations with the research team to agree on the form of project deliverables, evaluation indicators, and assessment methods. Projects for which a consensus cannot be reached through consultation will be considered as having forfeited the project approval.
• Projects evaluated as "excellent" may qualify for additional second-phase funding support, with a maximum of one project per direction eligible for this additional support. The evaluation method and requirements will be stipulated during the contract signing stage. For Topic 1, the second-phase funding is a fixed amount of RMB 1,000,000, and for Topic 2, the second-phase funding is a fixed amount of 250,000 yuan.

Important Dates: