Neurosymbolic Router Improves Reasoning Reliability on Edge Hardware

A neurosymbolic router uses a learned Deterministic Finite Automaton (DFA) to dispatch queries between small language models and exact symbolic solvers. This approach avoids forcing probabilistic models to approximate deterministic tasks like arithmetic or formal logic, reducing energy consumption on edge devices.