CIViC-Fact: a proof-of-concept framework for AI-assisted verification of cancer variant interpretations

CIViC-Fact reveals that less than 30% of cancer variant claims can be fully validated from abstracts alone, emphasizing the need for full-text evaluation in biomedical verification.

• Why it matters: Accurate and timely interpretation of genomic variants is essential for precision oncology, but current verification processes are slow and rely heavily on expert review, creating bottlenecks in clinical application.
• What they did: Developed as an expert-curated, full-text benchmark, CIViC-Fact links cancer-variant claims to detailed evidence from source publications, enabling standardized evaluation of AI systems in biomedical claim verification.
• The result: The framework successfully supports automated evidence retrieval for most new CIViC entries and facilitates ongoing comparison of retrieval models, promoting reproducible, lightweight updates in a rapidly evolving field.