Generalist VLMs Match Specialist Performance in Medical Imaging

Fine-tuned generalist vision-language models achieve comparable or superior performance to medical specialists, particularly on rare or out-of-distribution modalities. This suggests generalist architectures offer a scalable, cost-effective alternative to training massive, domain-specific medical models.