<p>A cognitive behavior framework is proposed to endow hierarchical geometric entities with human-like thinking and behaviors, aiming to shift assembly modeling—a critical digital product prototyping activity that involves mating part models together—from human-guided paradigm to autonomous paradigm, thus transferring the heavy cognitive burden of planning assembly task, identifying mating part models, reasoning mating regions, defining constraints, etc. from designers to entities themselves. The framework combines the previously developed concept of interaction feature pair (IFP), the belief-desire-intention (BDI) cognitive architecture, and the finite state machine (FSM) to establish cognitive behavior models for hierarchical geometric primitives, features, part models, and assembly models. IFP is self-evolved and utilized as a unified carrier of assembly intent (how an entity is expected to mate with other entities) in a designer’s brain and geometric entities’ cognitive behaviors that realize the intent. BDI takes advantage of IFP’s states to model thinking activities of geometric entities, making them autonomously build beliefs, update desires and determine intentions for assembly modeling. FSM controls the implementation of actions in entities’ intentions. An autonomous assembly modeling method is developed, supported by key algorithms for matching different types of IFPs and adjusting part model mating sequences. Two assembly modeling cases of a reducer and a Mecanum wheel are studied. Reductions in mouse operations (22% and 59%) and modeling time (47% and 51%) for the two cases, respectively, indicate that the cognitive behavior framework can support autonomous assembly modeling effectively and lift cognitive burden of designers greatly.</p>