OctoMed/MedXpertQA-Text

---
license: apache-2.0
dataset_info:
features:
- name: id
dtype: string
- name: question
dtype: string
- name: options
sequence: string
- name: answer
dtype: string
- name: medical_task
dtype: string
- name: body_system
dtype: string
- name: question_type
dtype: string
splits:
- name: test
num_examples: 2450
configs:
- config_name: default
data_files:
- split: test
path: data/test-*
---

# MedXpertQA-Text - Medical Expert QA (Text-Only)

## Description
This dataset contains expert-level medical questions in a text-only format (no images). Questions are designed to test deep medical knowledge and reasoning across various specialties, with 10 multiple choice options per question. We greatly appreciate and build from the original data source available at https://medxpertqa.github.io. We modify the format slightly to have `question`, `options`, and `answer` fields as described below.

## Data Fields
- `id`: Unique question identifier
- `question`: The medical question requiring expert knowledge
- `options`: Multiple choice answer options (10 choices, A–J)
- `answer`: The correct answer with option letter and text
- `medical_task`: Category of medical task (`Basic Science`, `Diagnosis`, `Treatment`)
- `body_system`: Anatomical body system relevant to the question
- `question_type`: Type of reasoning required (`Reasoning`, `Understanding`)

## Splits
- `test`: Test data for evaluation (2,450 examples)

## Usage
```python
from datasets import load_dataset

dataset = load_dataset("OctoMed/MedXpertQA-Text")
```

## Citation

If you find our work helpful, feel free to give us a cite!

```
@article{ossowski2025octomed,
title={OctoMed: Data Recipes for State-of-the-Art Multimodal Medical Reasoning},
author={Ossowski, Timothy and Zhang, Sheng and Liu, Qianchu and Qin, Guanghui and Tan, Reuben and Naumann, Tristan and Hu, Junjie and Poon, Hoifung},
journal={arXiv preprint arXiv:2511.23269},
year={2025}
}
```