Results for "standardized evaluation"

Running predictions on large datasets periodically.

Model optimizes objectives misaligned with human values.

Ensuring learned behavior matches intended objective.

Train/test environment mismatch.

Model behaves well during training but not deployment.

Required descriptions of model behavior and limits.

Governance of model changes.

AI applied to X-rays, CT, MRI, ultrasound, pathology slides.

Ability to correctly detect disease.

Failure to detect present disease.

US approval process for medical AI devices.