Frequently asked questions
What is Evals Review?
A free, curated directory of AI evals, benchmarks and rubrics. Each listing links to the original source, credits its creator and explains in plain English what it tests.
Who runs it?
Evals Review is an independent project run by Siddharth Mishra in India. It is not affiliated with any of the labs or creators listed.
How do you choose what to list?
We list evals that have a public, working source (a paper, repository, dataset or official page) and test something useful to AI builders. See the About & methodology page for details.
Are the ratings real?
Yes. We never invent ratings or seed fake reviews. Until real practitioners review an eval, its page says “No reviews yet” and shows only an editor note based on the source.
Who can write a review?
Anyone who signs in with GitHub, once reviews are switched on. One review per person per eval. Reviews must be about real experience using the eval.
Can sponsors buy better ratings?
No. Sponsored placements are always labelled as sponsored and never change ratings, reviews or editor notes.
How do I suggest an eval?
Use the Submit an eval page and include a link to the original source. An editor checks it before anything is listed.
An eval is listed wrongly or a link is broken. What do I do?
Use the Contact page. We correct factual errors as quickly as we can.
Is the newsletter free?
Yes. It's one short weekly email about new evals worth knowing. It uses double opt-in, and every email has an unsubscribe link.
Do you own the evals listed here?
No. All evals belong to their creators and come with their own licenses. Always check the license at the original source before using one.