Insights

Method documentation, result analysis, and guidance on interpreting results for your own caseload.

The archive

Showing 1–2 of 2 posts

The held-out set that wasn't

A fixed selection seed quietly turned our held-out set into a fixed exam every vendor could prepare for. How the bug works, and why a bigger paper is not the fix.

6 min read

Methodology

A refusal is not a safe answer

When a model declines a safety question we score it as though it never saw the item. Why a refusal earns no credit, and what that changes about the ranking.

7 min read

Scope. VetEval measures model performance on veterinary examination items. It is not clinical advice, and no result here licenses a model for clinical use.