RumorSphere: A Framework for Million-scale Agent-based Dynamic Simulation of Rumor Propagation

🤖 Introduction

RumorSphere is a novel dynamic and hierarchical social network simulation framework designed to model large-scale rumor propagation in social networks. RumorSphere constructs a simulated Twitter environment, where agents can perform actions such as posting, commenting, and retweeting. Through a multi-agent dynamic interaction strategy, we effectively combine agent-based models and LLMs, dynamically determining agent types in each round to improve simulation efficiency. The framework uses a hierarchical collaborative network to build more realistic relationship topologies. Additionally, RumorSphere offers counterfactual experimentation capabilities to explore various hypothetical scenarios, providing valuable insights for policymaking and public opinion management.

💡 Features

• Scalability: Supports simulations with millions of agents, dynamically adjusting to large-scale networks.

• Counterfactual Experiments: Simulate hypothetical scenarios and causal relationships for intervention strategies.

• Multi-agent Dynamic Interaction Strategy: : Differentiates between core and ordinary agents dynamically for efficient simulation.

• Hierarchical Collaborative Network: Enhances local community structures, fostering faster rumor propagation.

🚀 Getting Started

Requirements and Installation

Make sure you have Python >= 3.9

1. Clone the repository:

   git clone <repository-url>

2. Install Conda environment：

   conda create -n RumorSphere python=3.9
   conda activate RumorSphere
   pip install -e .

3. Set your OpenAI API key:

   # Export your OpenAI API key
   export OPENAI_API_BASE="your_api_base_here"
   export OPENAI_API_KEY="your_api_key_here"

4. You can adjust other parameters in agentverse/tasks/simulation/config.yaml.

Running the Simulation

CLI Demo

To start a simulation using the Command Line Interface (CLI), execute the following command:

agentverse-simulation --task simulation --ckpt logs/simulation.log

This command will initialize the simulation with the configuration found in agentverse/tasks/simulation/config.yaml, output the messages to messages.json, and log the details of the simulation in simulation.log.

🤝 Contributing

We welcome contributions from the community. If you'd like to contribute, please fork the repository and use a feature branch. Pull requests are warmly welcome.

📝 License

RumorSphere is distributed under the MIT License. You are free to use and modify it for your own projects.

📄 Cite

More details will be released soon.