On Generative Agents in Recommendation
Agent4Rec is a recommender system simulator featuring 1,000 LLM-empowered generative agents. Initialized from the MovieLens-1M dataset, these agents embody diverse social traits and preferences.
Key features:
- LLM-Empowered Agents: Simulates human-like behavior in recommendation environments using generative agents.
- Diverse Agent Personalities: Agents are initialized with varied social traits and preferences from the MovieLens-1M dataset.
- Interactive Simulation: Agents interact with personalized movie recommendations, performing actions like watching, rating, and evaluating.
- Customizable Recommender Systems: Supports various recommendation algorithms, including Random, Pop, MF, MultVAE, and LightGCN.
- Parallel Execution: Offers parallel execution mode to speed up simulations.
- Detailed Logging: Records interaction history for each agent, enabling in-depth analysis.
Use Cases:
- Evaluating the effectiveness of different recommendation algorithms.
- Studying the impact of user behavior on recommendation systems.
- Exploring the potential of LLM-empowered agents in simulating real-world interactions.