Reading your result
The index, the interval, the category breakdown, and which differences are real.
A result is one number with an interval around it, plus the breakdown that produced it. The interval is not decoration: it is the part that tells you whether a difference between two models means anything.
The index
The VetEval Index is a weighted percentage across the six competency categories and the safety component. The weights are the same for every model and are published on the methodology page.
The confidence interval
Every published score carries one. When two intervals overlap, the honest reading is that the two models performed comparably, whatever the ranking says about which is one place higher.
The breakdown
The per-category view is the useful part for improving a model. A strong overall index with a weak pharmacology score is a specific and actionable finding, and it is the sort of thing an aggregate hides.
What you see on your own runs
- The index, the interval and the category breakdown.
- A per-item view: the category, whether the answer was correct or declined, and the letter chosen.
- The safety status, and any flags raised.