xuhuang87's picture
update README.md
dfa0354
---
license: cc-by-4.0
task_categories:
- translation
language:
- en
- zh
- es
- fr
- de
- ru
- ja
- th
- sw
- te
- bn
- ar
- ko
- vi
- cs
- hu
- sr
multilinguality:
- multilingual
configs:
- config_name: arenahard_en
data_files: arenahard/arenahard_en.jsonl
- config_name: arenahard_zh
data_files: arenahard/arenahard_zh.jsonl
- config_name: arenahard_es
data_files: arenahard/arenahard_es.jsonl
- config_name: arenahard_fr
data_files: arenahard/arenahard_fr.jsonl
- config_name: arenahard_de
data_files: arenahard/arenahard_de.jsonl
- config_name: arenahard_ru
data_files: arenahard/arenahard_ru.jsonl
- config_name: arenahard_ja
data_files: arenahard/arenahard_ja.jsonl
- config_name: arenahard_th
data_files: arenahard/arenahard_th.jsonl
- config_name: arenahard_bn
data_files: arenahard/arenahard_bn.jsonl
- config_name: arenahard_sw
data_files: arenahard/arenahard_sw.jsonl
- config_name: arenahard_te
data_files: arenahard/arenahard_te.jsonl
- config_name: arenahard_ar
data_files: arenahard/arenahard_ar.jsonl
- config_name: arenahard_ko
data_files: arenahard/arenahard_ko.jsonl
- config_name: arenahard_vi
data_files: arenahard/arenahard_vi.jsonl
- config_name: arenahard_cs
data_files: arenahard/arenahard_cs.jsonl
- config_name: arenahard_hu
data_files: arenahard/arenahard_hu.jsonl
- config_name: arenahard_sr
data_files: arenahard/arenahard_sr.jsonl
- config_name: gpqa_en
data_files: gpqa/gpqa_en.jsonl
- config_name: gpqa_zh
data_files: gpqa/gpqa_zh.jsonl
- config_name: gpqa_es
data_files: gpqa/gpqa_es.jsonl
- config_name: gpqa_fr
data_files: gpqa/gpqa_fr.jsonl
- config_name: gpqa_de
data_files: gpqa/gpqa_de.jsonl
- config_name: gpqa_ru
data_files: gpqa/gpqa_ru.jsonl
- config_name: gpqa_ja
data_files: gpqa/gpqa_ja.jsonl
- config_name: gpqa_th
data_files: gpqa/gpqa_th.jsonl
- config_name: gpqa_bn
data_files: gpqa/gpqa_bn.jsonl
- config_name: gpqa_sw
data_files: gpqa/gpqa_sw.jsonl
- config_name: gpqa_te
data_files: gpqa/gpqa_te.jsonl
- config_name: gpqa_ar
data_files: gpqa/gpqa_ar.jsonl
- config_name: gpqa_ko
data_files: gpqa/gpqa_ko.jsonl
- config_name: gpqa_vi
data_files: gpqa/gpqa_vi.jsonl
- config_name: gpqa_cs
data_files: gpqa/gpqa_cs.jsonl
- config_name: gpqa_hu
data_files: gpqa/gpqa_hu.jsonl
- config_name: gpqa_sr
data_files: gpqa/gpqa_sr.jsonl
- config_name: humaneval_en
data_files: humaneval/humaneval_en.jsonl
- config_name: humaneval_zh
data_files: humaneval/humaneval_zh.jsonl
- config_name: humaneval_es
data_files: humaneval/humaneval_es.jsonl
- config_name: humaneval_fr
data_files: humaneval/humaneval_fr.jsonl
- config_name: humaneval_de
data_files: humaneval/humaneval_de.jsonl
- config_name: humaneval_ru
data_files: humaneval/humaneval_ru.jsonl
- config_name: humaneval_ja
data_files: humaneval/humaneval_ja.jsonl
- config_name: humaneval_th
data_files: humaneval/humaneval_th.jsonl
- config_name: humaneval_bn
data_files: humaneval/humaneval_bn.jsonl
- config_name: humaneval_sw
data_files: humaneval/humaneval_sw.jsonl
- config_name: humaneval_te
data_files: humaneval/humaneval_te.jsonl
- config_name: humaneval_ar
data_files: humaneval/humaneval_ar.jsonl
- config_name: humaneval_ko
data_files: humaneval/humaneval_ko.jsonl
- config_name: humaneval_vi
data_files: humaneval/humaneval_vi.jsonl
- config_name: humaneval_cs
data_files: humaneval/humaneval_cs.jsonl
- config_name: humaneval_hu
data_files: humaneval/humaneval_hu.jsonl
- config_name: humaneval_sr
data_files: humaneval/humaneval_sr.jsonl
- config_name: ifeval_en
data_files: ifeval/ifeval_en.jsonl
- config_name: ifeval_zh
data_files: ifeval/ifeval_zh.jsonl
- config_name: ifeval_es
data_files: ifeval/ifeval_es.jsonl
- config_name: ifeval_fr
data_files: ifeval/ifeval_fr.jsonl
- config_name: ifeval_de
data_files: ifeval/ifeval_de.jsonl
- config_name: ifeval_ru
data_files: ifeval/ifeval_ru.jsonl
- config_name: ifeval_ja
data_files: ifeval/ifeval_ja.jsonl
- config_name: ifeval_th
data_files: ifeval/ifeval_th.jsonl
- config_name: ifeval_bn
data_files: ifeval/ifeval_bn.jsonl
- config_name: ifeval_sw
data_files: ifeval/ifeval_sw.jsonl
- config_name: ifeval_te
data_files: ifeval/ifeval_te.jsonl
- config_name: ifeval_ar
data_files: ifeval/ifeval_ar.jsonl
- config_name: ifeval_ko
data_files: ifeval/ifeval_ko.jsonl
- config_name: ifeval_vi
data_files: ifeval/ifeval_vi.jsonl
- config_name: ifeval_cs
data_files: ifeval/ifeval_cs.jsonl
- config_name: ifeval_hu
data_files: ifeval/ifeval_hu.jsonl
- config_name: ifeval_sr
data_files: ifeval/ifeval_sr.jsonl
- config_name: lcb_v4_en
data_files: lcb_v4/lcb_v4_en.jsonl
- config_name: lcb_v4_zh
data_files: lcb_v4/lcb_v4_zh.jsonl
- config_name: lcb_v4_es
data_files: lcb_v4/lcb_v4_es.jsonl
- config_name: lcb_v4_fr
data_files: lcb_v4/lcb_v4_fr.jsonl
- config_name: lcb_v4_de
data_files: lcb_v4/lcb_v4_de.jsonl
- config_name: lcb_v4_ru
data_files: lcb_v4/lcb_v4_ru.jsonl
- config_name: lcb_v4_ja
data_files: lcb_v4/lcb_v4_ja.jsonl
- config_name: lcb_v4_th
data_files: lcb_v4/lcb_v4_th.jsonl
- config_name: lcb_v4_bn
data_files: lcb_v4/lcb_v4_bn.jsonl
- config_name: lcb_v4_sw
data_files: lcb_v4/lcb_v4_sw.jsonl
- config_name: lcb_v4_te
data_files: lcb_v4/lcb_v4_te.jsonl
- config_name: lcb_v4_ar
data_files: lcb_v4/lcb_v4_ar.jsonl
- config_name: lcb_v4_ko
data_files: lcb_v4/lcb_v4_ko.jsonl
- config_name: lcb_v4_vi
data_files: lcb_v4/lcb_v4_vi.jsonl
- config_name: lcb_v4_cs
data_files: lcb_v4/lcb_v4_cs.jsonl
- config_name: lcb_v4_hu
data_files: lcb_v4/lcb_v4_hu.jsonl
- config_name: lcb_v4_sr
data_files: lcb_v4/lcb_v4_sr.jsonl
- config_name: mgsm_en
data_files: mgsm/mgsm_en.jsonl
- config_name: mgsm_zh
data_files: mgsm/mgsm_zh.jsonl
- config_name: mgsm_es
data_files: mgsm/mgsm_es.jsonl
- config_name: mgsm_fr
data_files: mgsm/mgsm_fr.jsonl
- config_name: mgsm_de
data_files: mgsm/mgsm_de.jsonl
- config_name: mgsm_ru
data_files: mgsm/mgsm_ru.jsonl
- config_name: mgsm_ja
data_files: mgsm/mgsm_ja.jsonl
- config_name: mgsm_th
data_files: mgsm/mgsm_th.jsonl
- config_name: mgsm_bn
data_files: mgsm/mgsm_bn.jsonl
- config_name: mgsm_sw
data_files: mgsm/mgsm_sw.jsonl
- config_name: mgsm_te
data_files: mgsm/mgsm_te.jsonl
- config_name: mgsm_ar
data_files: mgsm/mgsm_ar.jsonl
- config_name: mgsm_ko
data_files: mgsm/mgsm_ko.jsonl
- config_name: mgsm_vi
data_files: mgsm/mgsm_vi.jsonl
- config_name: mgsm_cs
data_files: mgsm/mgsm_cs.jsonl
- config_name: mgsm_hu
data_files: mgsm/mgsm_hu.jsonl
- config_name: mgsm_sr
data_files: mgsm/mgsm_sr.jsonl
- config_name: nexus_en
data_files: nexus/nexus_en.jsonl
- config_name: nexus_zh
data_files: nexus/nexus_zh.jsonl
- config_name: nexus_es
data_files: nexus/nexus_es.jsonl
- config_name: nexus_fr
data_files: nexus/nexus_fr.jsonl
- config_name: nexus_de
data_files: nexus/nexus_de.jsonl
- config_name: nexus_ru
data_files: nexus/nexus_ru.jsonl
- config_name: nexus_ja
data_files: nexus/nexus_ja.jsonl
- config_name: nexus_th
data_files: nexus/nexus_th.jsonl
- config_name: nexus_bn
data_files: nexus/nexus_bn.jsonl
- config_name: nexus_sw
data_files: nexus/nexus_sw.jsonl
- config_name: nexus_te
data_files: nexus/nexus_te.jsonl
- config_name: nexus_ar
data_files: nexus/nexus_ar.jsonl
- config_name: nexus_ko
data_files: nexus/nexus_ko.jsonl
- config_name: nexus_vi
data_files: nexus/nexus_vi.jsonl
- config_name: nexus_cs
data_files: nexus/nexus_cs.jsonl
- config_name: nexus_hu
data_files: nexus/nexus_hu.jsonl
- config_name: nexus_sr
data_files: nexus/nexus_sr.jsonl
---
## Dataset Sources
- **Paper**: BenchMAX: A Comprehensive Multilingual Evaluation Suite for Large Language Models
- **Link**: https://huggingface.co/papers/2502.07346
- **Repository**: https://github.com/CONE-MT/BenchMAX
## Dataset Description
BenchMAX_Domain_Translation is a dataset of [BenchMAX](https://arxiv.org/pdf/2502.07346), which evaluates the translation capability on specific domains.
We collect the domain multi-way parallel data from other tasks in BenchMAX, such as math data, code data, etc.
Each sample contains one or three human-annotated translations.
## Usage
```bash
git clone --recurse-submodules https://github.com/CONE-MT/BenchMAX.git
cd BenchMAX
pip install -r requirements.txt
cd tasks/translation
tasks=("ifeval" "gpqa" "lcb_v4" "mgsm" "humaneval" "nexus" "arenahard")
max_tokens_list=(512 3072 2048 1024 1024 512 3072)
for i in "${!tasks[@]}"; do
task=${tasks[$i]}
max_tokens=${max_tokens_list[$i]}
python generate_translation.py -s en -t zh,es,fr,de,ru,ja,th,sw,bn,te,ar,ko,vi,cs,hu,sr --task-name $task --model-name ${model} --infer-backend vllm --max-tokens ${max_tokens}
python generate_translation.py -s zh,es,fr,de,ru,ja,th,sw,bn,te,ar,ko,vi,cs,hu,sr -t en --task-name $task --model-name ${model} --infer-backend vllm --max-tokens ${max_tokens}
python evaluate_translation.py -s en -t zh,es,fr,de,ru,ja,th,sw,bn,te,ar,ko,vi,cs,hu,sr --task-name $task --model-name ${model} --metrics spBLEU
python evaluate_translation.py -s zh,es,fr,de,ru,ja,th,sw,bn,te,ar,ko,vi,cs,hu,sr -t en --task-name $task --model-name ${model} --metrics spBLEU
done
```
## Supported Languages
Arabic, Bengali, Chinese, Czech, English, French, German, Hungarian, Japanese, Korean, Serbian, Spanish, Swahili, Telugu, Thai, Russian, Vietnamese
## Citation
If you find our dataset helpful, please cite this paper:
```
@article{huang2025benchmax,
title={BenchMAX: A Comprehensive Multilingual Evaluation Suite for Large Language Models},
author={Huang, Xu and Zhu, Wenhao and Hu, Hanxu and He, Conghui and Li, Lei and Huang, Shujian and Yuan, Fei},
journal={arXiv preprint arXiv:2502.07346},
year={2025}
}
```