SUBMITTING A MODEL

Provenance and declarations

What you are claiming when you submit, the five categories, and which one makes an entry ineligible.

Every submission declares where the model came from. This is the load-bearing statement on the whole platform: it decides what the leaderboard is allowed to say about who built what.

DeclarationChoose it whenRanked?
We created and trained this modelYour organization trained the weights.Yes
Built from open weights we did not trainA fine-tune, distillation or merge. Name the base model.Yes, with the base shown beside it
Our system around a third-party modelRetrieval, prompting or agents on top of somebody else's model.Evaluated as a system, with the underlying model named
Another provider's model, served through our APIReselling a model materially unchanged.No
Evaluated by VetEvalA baseline entry we ran ourselves. Staff only.Yes, with no claim of authorship

Why reselling is refused

If an unchanged model can be entered under a reseller's name, the leaderboard ranks somebody else's model under your brand, and a reader comparing two entries is comparing the same model twice. That is not a policy about fairness, it is about whether the board means anything.

Naming the base model

Every declaration except the first requires you to name what it is built on, and that name is shown publicly next to your result. A fine-tune that beats its base is a real achievement and the comparison is the interesting part.

Questions we get asked

Was this page helpful?

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