We introduce SMMILE, the first multimodal medical benchmark for evaluating in-context learning abilities of vision-language models.

New benchmark for studying complex scientific reasoning for language models. Particular focus on clinical genetics.