Commission a benchmark

Tell us what you want to find out, then choose who writes the keys and whether we check them. Each choice has its quality, and your benchmark's page says which you chose. The price is the same for everyone: what the run costs, plus 30% for our work.

About you

Your name Your email Your organisation (optional)

Your benchmark Its title What you want to find out What the questions are about (a few words, such as "cats and their care")

Who writes the keys

Every question has a key: how the matter stands, and what a sound answer contains. The judges read each answer against it.

I write them, or choose and amend from the questions our AI suggests ProlifAI writes and reviews them

Do we check them?

Automatic: it runs as soon as it is paid. The keys are not checked, so it runs under your name, marked as not checked. Reviewed: before it runs, we read every question and its key against the sources on the subject and correct what is wrong (a day or two), then send you the price. It may run under ProlifAI's name.

The models

Each question is asked three times to each model you choose, exactly as typed; three judges from other companies read every answer.

Loading the models…

Check every verse and hadith an answer cites against the texts (for questions on Islam)

What to measure

Keep the measures that fit your subject, and add up to five of your own. Each of your own is a yes-or-no question the judges answer of every answer; the page shows how often it was yes for each model. The cost per answer is always shown.

Loading the measures…

Its name, and what is published

Under ProlifAI's name Under my own name, as "run by ProlifAI with the AICAB method"

Publish everything: the figures and every answer Publish the figures; keep the answers private Publish nothing; only my link opens it

The price

Anything else I accept the terms for benchmarks.

Return to benchmarks