A behavioral taxonomy is the organized vocabulary for talking about what a model does — not just a list of behaviors, but a structure that shows how they relate. At the top level might be broad categories like “helpful behaviors,” “safety behaviors,” and “communication behaviors”; within each, more specific categories; within those, specific behaviors with examples. A well-structured taxonomy lets teams talk precisely about model behavior, build evaluation coverage systematically, and identify gaps or redundancies. For behavior architects, developing a behavioral taxonomy for their model is often an early foundational exercise: before you can measure behavior consistently or design training data deliberately, you need a shared language for what behaviors you’re trying to shape.