We've improved upon our current model for identifying explicit content, and in turn have reduced the number of false positives in GoGuardian Admin's Smart Alerts by 25%
This new model has shown that:
- The new X3 model improves precision to 94%, corresponding to an estimated 25% reduction of false positives.
- The new X3 model improves recall to 50%, capturing an estimated twice as many explicit pages as the current model does.
This improvement is a backend change. Customers will simply notice less false positives.