Skip to content

Provenance, licences and the decision record

Audit AI benchmark claims before you buy. Lesson 4, 21 min.

The last step turns your checks into a record that another person can audit. Ask for the run Ask the vendor for the run behind each claim: - Benchmark name, version, task IDs and source URL. - Exact model ID, date and parameters. - Prompt templates, tools and retry rule. - Grader version and rubric.…

The rest of this lesson is for learners enrolled in the course. See how Audit AI benchmark claims before you buy works

Published by . How we check each benchmark.

The benchmark newsletter

Updates on benchmarks and reported scores.

We only use your email to send this newsletter. Privacy policy.