MBZUAI/BiMed-V-1.6M

---
license: cc-by-nc-sa-4.0
task_categories:
- visual-question-answering
- image-to-text
language:
- en
- ar
multilinguality:
- multilingual
size_categories:
- 1M<n<10M
pretty_name: BiMed-V-1.6M
tags:
- medical
- biomedical
- multimodal
- vision-language
- instruction-tuning
- arabic
- bilingual
configs:
- config_name: stage1
data_files:
- split: train
path: BiMed-V_stage1.json
- config_name: stage2
data_files:
- split: train
path: BiMed-V_stage2.json
dataset_info:
features:
- name: id
dtype: string
- name: image
dtype: string
- name: conversations
list:
- name: from
dtype: string
- name: value
dtype: string
- name: language
dtype: string
- name: modality
dtype: string
splits:
- name: stage1
num_examples: 1691407
- name: stage2
num_examples: 467147
---

# BiMed-V-1.6M Dataset

[](https://github.com/mbzuai-oryx/BiMediX2)
[](https://arxiv.org/abs/2412.07769)
[](https://huggingface.co/collections/MBZUAI/bimedix2-675ee7528464dfd03f746127)
[](https://github.com/mbzuai-oryx/BiMediX/blob/main/LICENSE.txt)

## Dataset Description

**BiMed-V** is a comprehensive **Arabic-English multimodal bilingual instruction dataset** comprising over **1.6M instructions** designed for training medical Large Multimodal Models (LMMs). This dataset was curated as part of the BiMediX2 project, the first bilingual (Arabic-English) Bio-Medical Expert LMM for diverse medical modalities.

The dataset supports both English and Arabic languages, enabling the development of multilingual medical AI systems.

## Dataset Structure

The dataset is split into two training stages:

### Files

- **BiMed-V_stage1.json**: Pretraining data for vision-language projection alignment
- **BiMed-V_stage2.json**: Instruction finetuning data for task-specific adaptation

### Data Fields

Each sample in the dataset contains:
- **id**: Unique identifier for the sample
- **image**: Path or reference to the associated medical image (None for text-only instruction samples)
- **conversations**: List of conversation turns with roles
- **language**: Arabic or English
- **modality**: Text or Vision

## Citation

If you use BiMed-V dataset in your research, please cite:

```bibtex
@misc{mullappilly2024bimedix2biomedicalexpertlmm,
title={BiMediX2: Bio-Medical EXpert LMM for Diverse Medical Modalities},
author={Sahal Shaji Mullappilly and Mohammed Irfan Kurpath and Sara Pieri and Saeed Yahya Alseiari and Shanavas Cholakkal and Khaled Aldahmani and Fahad Khan and Rao Anwer and Salman Khan and Timothy Baldwin and Hisham Cholakkal},
year={2024},
eprint={2412.07769},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2412.07769},
}
```

## License

BiMed-V dataset is released under the **CC-BY-NC-SA 4.0 License**. For more details, please refer to the [LICENSE](https://github.com/mbzuai-oryx/BiMediX/blob/main/LICENSE.txt) file.

## Ethical Considerations

⚠️ **Important Notice**: This dataset is intended for **research purposes only** and is **not ready for clinical or commercial use**.

Users should be aware that:
- Medical AI models trained on this data should be validated by qualified healthcare professionals
- The dataset should not be used as the sole basis for medical diagnoses or treatment decisions
- Outputs from models trained on this data may contain errors, biases, or hallucinations
- Always verify AI-generated medical information with licensed healthcare providers