wildvision-eu (Basque Translation)

📚 Overview

wildvision-eu is a Basque-language translated dataset of the original WildVision/wildvision-bench multimodal question-answering dataset. A random sample of 5 000 English QA pairs was translated into Basque with HiTZ/Latxa-Llama-3.1-70B-Instruct.

Important: This is not the official dataset. It is an independent community translation created to support Basque-speaking researchers and practitioners.

✍️ Authors & Acknowledgements

• Original dataset: WildVision/wildvision-bench
• Basque translation & curation: <Lukas Arana / HiTZ>, 2025
  • Automatic translation with Latxa-Llama-70B

If you use this Basque split, please cite both the original WildVision dataset and this translation.

The JSONL schema matches the English source:

1. image_path: Image related to the instruction
2. instruction: Question related to the image
3. question_id: Unique identifier for the sample
4. response: Response given by a MLLM based on Latxa/Llama. These responses should not be taken as ground truth.

🔧 How We Built It

2. Machine translation – HiTZ/Latxa-Llama-3.1-70B-Instruct

No images were added or removed.

🚦 Limitations & Ethical Considerations

• Non-official: The WildVision team has not reviewed or endorsed this edition; meaning shifts may remain.
• Model biases: Latxa-Llama-70B can introduce or amplify biases present in either the source data or the MT system.

💻 Quick Start

from datasets import load_dataset

ds = load_dataset(
"your-hf-org/wildvision-benchmark-eu",
split="train"
)

print(ds)

All fields mirror the English original; only textual content is localized.

📜 License

This translation inherits the same license as the upstream WildVision dataset (CC-BY-SA-4.0). By downloading or using these files, you agree to comply with that license, including proper attribution to both the WildVision authors and the Basque translation contributors.

🏷️ Citation

@misc{lu2024wildvisionevaluatingvisionlanguagemodels, title={WildVision: Evaluating Vision-Language Models in the Wild with Human Preferences}, author={Yujie Lu and Dongfu Jiang and Wenhu Chen and William Yang Wang and Yejin Choi and Bill Yuchen Lin}, year={2024}, eprint={2406.11069}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2406.11069}, }

dataset_info: features: - name: image_path dtype: image - name: instruction dtype: string - name: question_id dtype: string - name: response dtype: string splits: - name: eu num_bytes: 137431474.0 num_examples: 500 - name: latxa_responses num_bytes: 137810849.0 num_examples: 500 - name: llama_responses num_bytes: 137874574.0 num_examples: 500 download_size: 404643720 dataset_size: 413116897.0 configs: - config_name: default data_files: - split: eu path: data/eu-* - split: latxa_responses path: data/latxa_responses-* - split: llama_responses path: data/llama_responses-*