As a patroller,
I'd like to have revertrisk language-agnostic predictions exposed via recent changes filters so that they can further assist my work while trying to fight vandalism.
Exposing revertrisk predictions as separate filters can act complementary to the existing damaging and goodfaith models but more importantly will allow us to enable the ORES extension in more wikis/languages for which we don't have such models at the moment.
We also need to update https://www.mediawiki.org/wiki/Help:New_filters_for_edit_review to document the addition of new filters.
Per T348298#9559718 we will start with one filter level at 0.95