MedQA-USMLE: Medical Question Answering Dataset from USMLE Exams

A large-scale open domain question answering dataset introduced by Jin et al. in 2020, derived from medical licensing exams. The dataset, hosted on HuggingFace, is designed for evaluating AI models on clinical reasoning. It was last updated on the platform in January 2023.

Use Cases

• Benchmarking medical question-answering models based on USMLE-style multiple-choice questions.
• Training language models for clinical reasoning based on open-domain medical exam content.
• Evaluating the factual knowledge and diagnostic reasoning of AI systems based on exam-derived questions.

Strengths

• Dataset is based on a peer-reviewed publication from 2020, providing academic credibility.
• Platform tags indicate a 'large-scale' dataset, likely containing a substantial number of questions.
• Data is in English and focused on US medical licensing exams (USMLE), providing a standardized domain.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and exact size are unknown, which may limit suitability assessment.
• Last updated 2023-01-24; freshness should be verified against the original source.

Provenance

Source

Original dataset introduced by Jin et al. in the paper 'What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams'.

Collection Method

Derived from medical licensing exams, likely the United States Medical Licensing Examination (USMLE).

Freshness

Last updated on the platform: 2023-01-24 19:18:09.

Geography

Platform tags indicate a 'us' region, suggesting a focus on United States medical exams.