MODE: Mutual Optimality in Direct Effects of Reciprocal Recommendations in Matching Markets

This repository contains the implementation code for experiments with synthetic data in the paper "MODE: Mutual Optimality in Direct Effects of Reciprocal Recommendations in Matching Markets" by Yoji Tomita (CyberAgent, AI Lab), RecSys'26.

Setup

If you can use uv, you can set up the environment with:

uv sync

Otherwise, you need following packages:

• Python 3.12
• Numpy 2.4.4
• CVXPY 1.8.2
• PyTorch 2.11.0

Usage

You can run the main experiment script with uv:

uv run main.py --n 10 --exam_type inv --pref_lambda 0.8 --seed 0

or with your own python environment:

python main.py --n 10 --exam_type inv --pref_lambda 0.8 --seed 0

The options are:

• --n or -n: Number of jobs. Number of candidates is set to 1.5*n. Default is 10.
• --exam_type or -e: Type of examination probability vectors. Choices are "log", "inv", and "exp". Default is "inv".
• --pref_lambda or -l: Lambda parameter for preference generation. Default is 0.8.
• --seed or -s: Random seed for preference generation. Default is 0.
• --nonverbose or -nv: Disable verbose output. By default, verbose output is enabled.
• --result_file_path or -r: Path to save the result JSON file. If not provided, the result will be saved to a default file name based on the parameters.

Files

• main.py: Main script to run the experiment.
• market.py: Contains the Market class which defines the matching market.
• naive.py: Contains the implementation of the Naive method.
• reciprocal.py: Contains the implementation of the Reciprocal method.
• tu.py: Contains the implementation of the TU method.
• approx_sw.py: Contains the implementation of the ApproxSW method.
• direct_sw.py: Contains the implementation of the DirectSW method.
• mode.py: Contains the implementation of the MODE method.

Citation

If you find our work useful in your research, please consider citing:

@inproceedings{tomita2026mode,
  title={MODE: Mutual Optimality in Direct Effects of Reciprocal Recommendations in Matching Markets},
  author={Yoji Tomita},
  booktitle={Proceedings of the 20th ACM Conference on Recommender Systems},
  year={2026}
}

License

This repository is licensed under the MIT License.