Health AI papers often omit how their agents are built and governed
A preprint scores 279 health-AI records, finding evaluation details were more complete than architecture, governance and reproducibility.
Developing Light ยท https://developinglight.com/editorial/developing-light