CynapsClinical Benchmarks

Physician review

Help build a clinical AI benchmark for Indian medicine

Practising specialists review every question and grading criterion before any AI model is tested on them.

What you will do
Judge whether each question is clinically sound and each grading criterion is a fair basis for marking. Write questions of your own if you wish.
Your time
Each review shows its size and a time estimate before you begin. Work at your own pace; every mark is saved as you go.
Credit
Credited by name if you choose. You see the final wording of anything that will carry your name before it is published.
Confidentiality
The items are unpublished. The participation agreement asks you to keep them confidential until the benchmark is published.

Declared interest

This benchmark is built and funded by nBall. Its questions and grading criteria were drafted by an AI system developed by nBall, working from verified clinical sources. At this stage, specialists are reviewing the benchmark itself: whether each question is clinically sound and each criterion is a valid basis for grading. Once finalised, the benchmark will be used to evaluate frontier AI models, with Cynaps assessed alongside them. Because the developer has an interest in the performance of one of the systems evaluated, independent specialist review is an essential safeguard against bias.

CynapsClinical Benchmarks

Help build a clinical AI benchmark for Indian medicine

Practising specialists review every question and grading criterion before any AI model is tested on them.

Step 1 of 2 · Email

Sign in

Enter the address your invitation was sent to and we will email you a six digit code.

Reviewers join by invitation. If you were expecting one and it has not arrived, check your spam folder or ask the person who invited you.

What you will do
Judge whether each question is clinically sound and each grading criterion is a fair basis for marking. Write questions of your own if you wish.
Your time
Each review shows its size and a time estimate before you begin. Work at your own pace; every mark is saved as you go.
Credit
Credited by name if you choose. You see the final wording of anything that will carry your name before it is published.
Confidentiality
The items are unpublished. The participation agreement asks you to keep them confidential until the benchmark is published.

Declared interest

This benchmark is built and funded by nBall. Its questions and grading criteria were drafted by an AI system developed by nBall, working from verified clinical sources. At this stage, specialists are reviewing the benchmark itself: whether each question is clinically sound and each criterion is a valid basis for grading. Once finalised, the benchmark will be used to evaluate frontier AI models, with Cynaps assessed alongside them. Because the developer has an interest in the performance of one of the systems evaluated, independent specialist review is an essential safeguard against bias.

Cynaps