I ran some experiments using only the core of the Community Notes algorithm, taking votes to be helpfulness ratings. I didn’t get anything super interesting out of it, though I might have had implementation bugs. The top posts according to the model seemed fine, and then I didn’t allocate much time to poking around at it any more.
I ran some experiments using only the core of the Community Notes algorithm, taking votes to be helpfulness ratings. I didn’t get anything super interesting out of it, though I might have had implementation bugs. The top posts according to the model seemed fine, and then I didn’t allocate much time to poking around at it any more.