Expert Medical Question-Answering Benchmark With Multimodal Tasks

MedXpertQA is a benchmark dataset containing 4,460 questions for evaluating expert-level medical knowledge and reasoning. It was created by TsinghuaC3I and features both text-based and multimodal tasks that integrate structured clinical data with images. The dataset was last updated in July 2025.

Use Cases

• Benchmarking model performance on expert-level medical questions across diverse specialties and tasks.
• Training multimodal models to answer questions using structured clinical information and medical images.
• Evaluating advanced reasoning capabilities on text-based medical question-answering tasks.
• Analyzing model performance across different medical body systems and image types featured in the dataset.

Strengths

• Contains 4,460 questions for evaluation.
• Spans diverse medical specialties, tasks, and body systems.
• Includes a multimodal subset integrating structured data with images.

Limitations

• Specific sample size, column details, and file formats are not provided.
• Potential for geographic or institutional bias in the source of medical questions is unknown.

Provenance

Source

TsinghuaC3I.

Freshness

Last updated on 2025-07-09.