OmniMedVQA-V2: Medical Visual Question Answering Benchmark Across 12 Modalities

OmniMedVQA-V2 is a large-scale medical visual question answering benchmark covering 12 imaging modalities and 5 clinical question types. The v2 release introduces 13 granular named configurations for modalities and question types, with train/test partitions following the Med-R1 standard. Images are sourced from the canonical foreverbeliever/OmniMedVQA release, with restricted-access images excluded.

Use Cases

• Benchmarking medical VQA model performance based on the 12 specified imaging modalities.
• Training multimodal models on a diverse set of clinical question types described in the dataset.
• Evaluating model generalization across granular dataset configurations (mod-* and qt-*).
• Studying the intersection of medical imaging and natural language processing for clinical applications.

Strengths

• Covers 12 distinct medical imaging modalities, providing broad coverage.
• Includes 5 clinical question types, enabling diverse task evaluation.
• Uses standardized Med-R1 partitioning for train/test splits.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Images sourced from the canonical foreverbeliever/OmniMedVQA release.

Freshness

Last updated 2026-04-18 20:34:07; freshness should be verified.