Evals

MAST evaluations go beyond standard accuracy metrics. Each benchmark in the suite measures a distinct dimension of clinical competence, and all evaluation protocols are developed with board-certified physicians and validated against expert consensus.

Safety and Harm Avoidance

Measures whether AI-generated recommendations could lead to patient harm. This includes detecting dangerous drug interactions, flagging unsafe dosages, and checking whether recommendations follow established clinical guidelines.

Diagnostic Accuracy

Measures the ability to arrive at correct diagnoses given a patient's symptoms, lab results, and imaging findings. Responses are scored against expert panel consensus.

Clinical Reasoning

Measures how well AI updates its clinical thinking when presented with new, ambiguous, or conflicting information. This reflects how doctors actually make decisions in practice.

Multimodal Comprehension

Measures the ability to interpret clinical images, pathology slides, and radiology findings alongside written patient data.

Suggest an Eval

Have an idea for a new safety test or clinical workflow check? We welcome proposals from researchers and clinicians.

Submit suggestion

Are You a Healthcare AI Company?

Want your product independently evaluated by board-certified clinicians? We offer private evaluation engagements and public leaderboard inclusion.

Get in touch ↗

All evaluations are blinded, clinician-governed, and methodologically independent.