All notes

Governance note

Evaluation access without pay-to-rank

How VetoBench can accept model access or credits without selling influence over the leaderboard.

01

The cold-start constraint

Video evaluation is expensive. A young benchmark cannot cover every production model by paying retail generation prices, and a provider can often supply the same inference at far lower internal cost.

Provider-funded access is therefore practical. It is also a conflict that must be visible and controlled rather than ignored.

02

What support may buy

A provider may supply a production-equivalent endpoint, a capped credit wallet or technical help required to run the published protocol. Support may also fund infrastructure that is shared across eligible models.

Support does not buy prompt selection, scoring changes, advance access to private holdouts, favourable placement or the right to suppress a completed public evaluation.

03

What will be disclosed

Each model page will identify the endpoint, model version, evaluation date, run settings and source of access. Provider-supplied access will be labelled. When resources permit, Capvise Labs will compare selected runs against independently funded or anonymous access to check for material differences.

Document status

This page records the lab's position at publication. It does not claim an unreleased model, completed benchmark run or peer-reviewed result. Material changes will be dated.