How we make a claim.

A figure is only worth publishing with the conditions that produced it and the cases where it failed. These are the rules every number on this site follows.

The pass mark comes first

Before a dataset is run, we write down what result would count as a pass and what would count as a failure. The result is judged against that, and the mark is never moved afterwards.

The engine produces the number

Results come from the Senua AI engine itself, not from a prototype that imitates it. Figures that came from prototype scripts have been withdrawn.

Public data where possible

Most results use public datasets that anyone can obtain and check. Internal datasets are labelled as such.