CAUM: Trajectory-Based Detection of Structural Stagnation

CAUM is a trajectory-based diagnostic framework for early detection of structural stagnation and loop-like behavior in agents navigating discrete or discretized state spaces. CAUM operates exclusively on observed state trajectories and does not inspect internal logic, code, parameters, or task semantics. The framework defines two complementary signals—Exploration/Novelty (E) and Friction/Re-derivation (D)—whose joint dynamics characterize regimes of healthy exploration, productive repetition, structural stagnation, and loop-like behavior. This record includes:(1) the core CAUM whitepaper,(2) an empirical and supplementary validation report across multiple domains, and(3) an authorship and first disclosure record intended to establish provenance and timestamp. The materials are provided for academic and technical reference.