JevMade

Sign in
← Back to experiments

Benchmarks & research

Jev Clinical Calibration

A pre-registered study compared Jev 1.13 against Claude Opus models on clinical diagnosis vignettes, finding Jev matched frontier models on short complaints but lagged on multi-symptom synthetic patients.

Source screenshot of Jev Clinical Calibration
SOURCE SCREENSHOTFull screenshot ↗

What it does

Some synthetic cases received zero probability for the true diagnosis, the authors report. They also found that uncertainty depended on whether an abstention option was offered. This preprint has not been peer-reviewed and must not guide clinical practice. Neither the synthetic results nor the short-vignette results have been reproduced here.

Keep this for later

Sign in to bookmark experiments, guides and videos, and keep notes only you can see.

Continue to sign in

We’ll bring you back to this listing.