[NAACL 2025] Is Translation All You Need? A Study on Solving Multilingual Tasks with Large Language Models

This repository contains the code and datasets for the paper "Is Translation All You Need? A Study on Solving Multilingual Tasks with Large Language Models", which was accepted at NAACL 2025.

Datasets

All the datasets are in datasets folder. The datasets also include the translations of the original data into English, which are used for the experiments in the paper. The datasets include:

• NLP tasks: MGSM, XCOPA, XNLI, XNLI, PAWS-X, MKQA, XL-sum
• Real-world queries: ShareGPT

Inference

Prepare the environment with yml file

conda env create -f environment.yml

Activate the environment

conda activate multilingual

To run the inference, you can update the parameters and use the following command:

source 01_run_NLP_tasks.sh
source 02_run_shareGPT.sh

Citation

@misc{liu_is_2024,
	title = {Is {Translation} {All} {You} {Need}? {A} {Study} on {Solving} {Multilingual} {Tasks} with {Large} {Language} {Models}},
	url = {http://arxiv.org/abs/2403.10258},
	publisher = {arXiv},
	author = {Liu, Chaoqun and Zhang, Wenxuan and Zhao, Yiran and Luu, Anh Tuan and Bing, Lidong},
	month = jun,
	year = {2024},
}