Metagente

This repository contains the source code implementation of metagente and the datasets used to replicate the experimental results

Getting Started

Dependencies

• Python version 3.10.12
• Python packages are listed in requirements.txt
• MongoDB

Running the application

Installing necessary packages:

pip install -r requirements.txt

Running the optimization code:

python main.py
--train_data_file data/train_data.csv
--train_result_dir result/train

Running the evaluation code:

python evaluation.py
--test_data_file data/test_data.csv
--test_result_dir result/test

Repositories structure

This repository contains the source code, datasets, and results for the experiments described in our paper. The structure of the project is as follows:

Data

This folder contains the input datasets used in the experiments.

• ES.csv: Main dataset used for the experiments.
• TS10.csv: Subset of the dataset used for testing with 10 samples.
• TS50.csv: Subset of the dataset used for testing with 50 samples.

Results

This folder contains the outputs generated during the experiments.

• GITSUM_TS10.txt: Results generated by the GITSUM model for the TS10 dataset.
• GITSUM_TS50.txt: Results generated by the GITSUM model for the TS50 dataset.
• LLAMA_TS10.csv: Summary results from the LLAMA model for the TS10 dataset.
• LLAMA_TS50.csv: Summary results from the LLAMA model for the TS50 dataset.
• METAGENTE: This folder contains parallel and sequential results from metagente for TS10 dataset. Moreover, Statistical tests are provided.

Tools

This folder contains the Python scripts used to run the experiments and generate results.

• GITSUM.py: Script for running the GITSUM model on the datasets.
• LLAMA_SUMMARY.py: Script for processing and summarizing results from the LLAMA model.