Datasets:
metadata
language:
- bg
- de
- en
- es
- fr
- hi
- pt
- ru
- zh
configs:
- config_name: en
data_files:
- split: test
path: en_mme.jsonl
- config_name: bg
data_files:
- split: test
path: bg_mme.json
- config_name: de
data_files:
- split: test
path: de_mme.json
- config_name: es
data_files:
- split: test
path: es_mme.json
- config_name: fr
data_files:
- split: test
path: fr_mme.json
- config_name: hi
data_files:
- split: test
path: hi_mme.json
- config_name: pt
data_files:
- split: test
path: pt_mme.json
- config_name: ru
data_files:
- split: test
path: ru_mme.json
- config_name: zh
data_files:
- split: test
path: zh_mme.json
This is the multilingual version of MME. If you find it useful, we hope you can cite our paper.
@article{ye2025claim,
title={CLAIM: Mitigating Multilingual Object Hallucination in Large Vision-Language Models with Cross-Lingual Attention Intervention},
author={Ye, Zekai and Li, Qiming and Feng, Xiaocheng and Qin, Libo and Huang, Yichong and Li, Baohang and Jiang, Kui and Xiang, Yang and Zhang, Zhirui and Lu, Yunfei and others},
journal={arXiv preprint arXiv:2506.11073},
year={2025}
}