Skip to content

Latest commit

 

History

History
64 lines (45 loc) · 2.57 KB

README.md

File metadata and controls

64 lines (45 loc) · 2.57 KB
Raw

Lusifer: LLM-based User Simulated Feedback Environment for Online Recommender Systems

Lusifer (LLM-based User SImulated Feedback Environment for online Recommender systems) is a platform designed to simulate user behavior and generate feedback for recommender systems using Large Language Models (LLMs). This environment helps researchers evaluate and improve their recommender systems by providing a scalable and adaptable framework for user simulation.

For more information, read the paper:
Lusifer: LLM-based User Simulated Feedback Environment for Online Recommender Systems

Getting Started

Set OpenAI Connection

from lusifer import Lusifer 
lusifer = Lusifer(users_df, items_df, ratings_df) 
lusifer.set_openai_connection(api_key='your_openai_api_key', model='gpt-3.5-turbo')

Set LLM Instructions and Prompts

lusifer.set_llm_instruction(instructions="your_llm_instructions") 
lusifer.set_prompts(
    prompt_summary="your_prompt_to_generate_summary", 
    prompt_update_summary="your_prompt_to_update_summary", 
    prompt_simulate_rating="your_prompt_to_simulate_rating" 
)

set column names

lusifer.set_column_names(user_feature="user_info",
                         item_feature="movie_info",
                         user_id="user_id",  # set by default
                         item_id="movie_id",
                         timestamp="timestamp",  # set by default
                         rating="rating")  # set by default

    
# you can set the prompts as below, or ignor this and use the default prompts
lusifer.set_prompts(prompt_summary, prompt_update_summary, prompt_simulate_rating)

# you can set the path to store intermediate storing procedure. By default, they will be saved on Root.
lusifer.set_saving_path(self, path="")

Run the Sample Experiment

A sample experiment on the MovieLens dataset is available: here

Access Sample Prompts, Datasets, and Outputs

Explore sample prompts, datasets, and outputs here. Contributions This is the first version of Lusifer. We welcome contributions and suggestions to improve and expand the capabilities of this tool.

References

Paper: Lusifer: LLM-based User Simulated Feedback Environment for Online Recommender Systems.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact

For any inquiries or further information, please contact Danial Ebrat.