README.md

metadata

dataset_info:
  features:
    - name: idx
      dtype: string
    - name: question
      dtype: string
    - name: options
      struct:
        - name: A
          dtype: string
        - name: B
          dtype: string
        - name: C
          dtype: string
        - name: D
          dtype: string
        - name: E
          dtype: string
    - name: answer
      dtype: string
    - name: explanation
      dtype: string
    - name: link
      dtype: string
  splits:
    - name: op4_test
      num_bytes: 1210993
      num_examples: 308
    - name: op5_test
      num_bytes: 1218063
      num_examples: 308
  download_size: 1206626
  dataset_size: 2429056
configs:
  - config_name: default
    data_files:
      - split: op4_test
        path: data/op4_test-*
      - split: op5_test
        path: data/op5_test-*
task_categories:
  - question-answering
tags:
  - medical,
  - clinical,
  - multiple-choice
  - usmle

Medbullets

HuggingFace upload of a multiple-choice QA dataset of USMLE Step 2 and Step 3 style questions sourced from Medbullets. If used, please cite the original authors using the citation below.

Dataset Details

Dataset Description

The dataset contains four splits:

• op4_test: four-option multiple-choice QA (choices A-D)
• op5_test: five-option multiple-choice QA (choices A-E)

op5_test contains the same content as op4_test, but with one additional answer choice to increase difficulty. Note that while the content is the same, the letter choice corresponding to the correct answer is sometimes different between these splits.

Dataset Sources

• Repository: https://github.com/HanjieChen/ChallengeClinicalQA
• Paper: https://arxiv.org/pdf/2402.18060v3

Direct Use

import json
from datasets import load_dataset, Dataset

def _strip_E(split):
    for ex in split:
        ex = dict(ex)
        ex["options"] = {k: v for k, v in ex["options"].items() if k != "E"}
        yield ex

if __name__ == "__main__":
    op4_test, op5_test = load_dataset("mkieffer/Medbullets", split=["op4_test", "op5_test"])

    # remove the "E" option from op4 split
    op4_test = Dataset.from_generator(lambda: _strip_E(op4_test))

    print("\nop4_test:\n", json.dumps(op4_test[0], indent=2))
    print("\nop5_test:\n", json.dumps(op5_test[0], indent=2))

Citation

@inproceedings{chen-etal-2025-benchmarking,
    title = "Benchmarking Large Language Models on Answering and Explaining Challenging Medical Questions",
    author = "Chen, Hanjie  and
      Fang, Zhouxiang  and
      Singla, Yash  and
      Dredze, Mark",
    editor = "Chiruzzo, Luis  and
      Ritter, Alan  and
      Wang, Lu",
    booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
    month = apr,
    year = "2025",
    address = "Albuquerque, New Mexico",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.naacl-long.182/",
    doi = "10.18653/v1/2025.naacl-long.182",
    pages = "3563--3599",
    ISBN = "979-8-89176-189-6",
    abstract = "LLMs have demonstrated impressive performance in answering medical questions, such as achieving passing scores on medical licensing examinations. However, medical board exams or general clinical questions do not capture the complexity of realistic clinical cases. Moreover, the lack of reference explanations means we cannot easily evaluate the reasoning of model decisions, a crucial component of supporting doctors in making complex medical decisions. To address these challenges, we construct two new datasets: JAMA Clinical Challenge and Medbullets. JAMA Clinical Challenge consists of questions based on challenging clinical cases, while Medbullets comprises simulated clinical questions. Both datasets are structured as multiple-choice question-answering tasks, accompanied by expert-written explanations. We evaluate seven LLMs on the two datasets using various prompts. Experiments demonstrate that our datasets are harder than previous benchmarks. In-depth automatic and human evaluations of model-generated explanations provide insights into the promise and deficiency of LLMs for explainable medical QA."
}