CriteriaBot

Users begin by writing criteria in everyday English, which can be grouped for repeated application across different evaluations. Content is then submitted for analysis, resulting in individual true or false outcomes per criterion that integrate easily into existing workflows for tasks such as filtering or categorization. The underlying process involves gathering supporting facts from external references before independent model assessments occur. These evaluations incorporate user-specific preferences to weight contributions, ultimately producing a single consolidated verdict while allowing for ongoing refinement through personal feedback. This approach supports diverse scenarios including moderation of discussions, detection of unwanted patterns, enforcement of stylistic guidelines, and organization of incoming data streams without reliance on any single model for final decisions.