YuLan-OneSim - A social simulator developed by the Renmin University Gaoling AI Team
YuLan-OneSim is a new type of social simulator developed by the RUC-GSAI team at Renmin University of China's Gaoling AI Institute. Based on Large Language Model (LLM) agents, it simulates human social behavior without requiring programming to build the model...
What is YuLan-OneSim?
YuLan-OneSim is a novel social simulator developed by the RUC-GSAI team at the Gaoling School of AI, Renmin University of China. Based on Large Language Model (LLM) agents, it simulates human social behavior without requiring programming to build simulation scenarios; code is generated through natural language interaction. YuLan-OneSim provides over 50 default scenarios covering eight major social science fields, supports large-scale simulations of up to 100,000 agents, and automatically optimizes the LLM based on external feedback. YuLan-OneSim also features AI social researcher functionality, automatically generating reports from research topics and completing the entire social science research cycle. YuLan-OneSim promotes the deep integration of social science and AI, providing a powerful tool for social science research.
Main functions of YuLan-OneSim
- Building simulation scenarios without programmingUsers can use natural language to describe and refine the simulated scenario, and the system will automatically generate the corresponding execution code, reducing the requirements for programming skills.
- A rich library of default scenesIt provides over 50 default simulation scenarios covering eight major social science fields, including economics, sociology, and political science, offering researchers a wealth of research materials.
- Evolvable simulationIt automatically optimizes LLM based on external feedback, improving the accuracy and reliability of simulations.
- Large-scale simulation capabilityBased on a distributed architecture, it supports the simulation of up to 100,000 agents.
- AI social researchersIt automatically transforms research topics into specific simulation scenarios and generates research reports, automating the research process from the proposal of research topics to the generation of reports.
The technical principle of YuLan-OneSim
- Scene formalizationBased on the Overview, Design Concepts, and Details (ODD) protocol, user requirements are transformed into structured scenario descriptions.
- Behavior graph constructionBased on the ODD protocol, extract Agent type and behavioral logic, generate behavioral graph, and define the interaction logic between Agents.
- Code generationGenerate executable simulation code based on behavior graphs, using a modular code generation method to ensure the accuracy and maintainability of the code.
- Scene SpecificationsGenerate environmental data, agent configuration data, and agent relationship data to ensure the integrity and consistency of the simulation.
- Simulation subsystemThe Agent consists of multiple modules, including configuration, memory, planning, and action modules, supporting high customization. It utilizes an event-driven asynchronous event bus to support parallel computing and efficient event processing. Based on a master-slave node architecture, it supports parallel computing for large-scale Agents, optimizing communication efficiency and resource allocation.
- Feedback-driven evolutionary subsystemBased on a multi-agent framework (Verifier–Reasoner–Refiner–Tuner, VR²T), simulation results are evaluated and optimized. The system-generated prompt responses are validated, inferred, corrected, and fine-tuned to improve the performance of LLM.
- AI Social Researcher SubsystemThe experimental design module transforms research topics into specific simulation scenarios, including generating candidate research questions, assessing scenario feasibility, and generating an ODD (Optical Design Principle). The report generation module generates detailed analysis reports based on the simulation results, including data interpretation, report structure generation, report writing, and report review.
YuLan-OneSim project address
- GitHub repository:https://github.com/RUC-GSAI/YuLan-OneSim
- arXiv technical paper:https://arxiv.org/pdf/2505.07581
Application scenarios of YuLan-OneSim
- Economic fieldSimulate market dynamics, such as labor matching, auction markets, and bank reserves, to analyze the impact of policies and mechanisms.
- SociologyThe study of social phenomena, such as cultural capital, information dissemination, and social norms, aims to understand the interaction between social structure and individual behavior.
- field of political scienceSimulate voter behavior, policy implementation, and electoral polarization to analyze the social impact of systems and policies.
- field of psychologySimulate cognitive dissonance, emotional contagion, and conformity to explore the underlying mechanisms of psychological behavior.
- public healthSimulate the spread of infectious diseases, health inequalities, and community health mobilization to evaluate the effectiveness of intervention measures.