Beyond Static Testbeds: An Interaction-Centric Agent Simulation Platform for Dynamic Recommender Systems

Evaluating and iterating upon recommender systems is crucial, yet traditional A/B testing is resource-intensive, and offline methods struggle with dynamic user-platform interactions. While agent-based simulation is promising, existing platforms often lack a mechanism for user actions to dynamically reshape the environment. To bridge this gap, we introduce RecInter, a novel agent-based simulation platform for recommender systems featuring a robust interaction mechanism. In RecInter platform, simulated user actions (e.g., likes, reviews, purchases) dynamically update item attributes in real-time, and introduced Merchant Agents can reply, fostering a more realistic and evolving ecosystem. High-fidelity simulation is ensured through Multidimensional User Profiling module, Advanced Agent Architecture, and LLM fine-tuned on Chain-of-Thought (CoT) enriched interaction data. Our platform achieves significantly improved simulation credibility and successfully replicates emergent phenomena like Brand Loyalty and the Matthew Effect. Experiments demonstrate that this interaction mechanism is pivotal for simulating realistic system evolution, establishing our platform as a credible testbed for recommender systems research.

🕹️ Environment Setup

1. Create a new virtual environment

conda create --name recinter python=3.10
conda activate recinter

2. Install requirement packages

pip install --upgrade pip setuptools

pip install -e .  # This will install dependencies as specified in pyproject.toml

3. Set Up Environment Variables

# Export your OpenAI API key
export OPENAI_API_KEY=<insert your OpenAI API key>
export OPENAI_API_BASE_URL=<insert your OpenAI API BASE URL>  #(Should you utilize an OpenAI proxy service, kindly specify this)

🔧 Run the Code

1. Run the main simulation

python simulation.py --config_path config/default.yaml # You can modify the parameters via the command line or configuration file

2. Run the main simulation with open-souce models

python deploy.py # Use vllm to deploy LLM lcoally
python simulation.py --config_path config/default.yaml --inference.model_type qwen-2.5

🌹 Acknowledgmentsons

This project use LLaMA-Factory for mdoels fine-tuning. Part of the code is borrowed from the OASIS and CAMEL repo. Special thanks for providing the foundation for this work.