CircuitGate: Logic-Consistent Circuit-Level Functional Modeling for And-Inverter Graphs
CircuitGate improves how computer circuits are represented by learning models called And-Inverter Graphs (AIGs). It moves beyond local gate analysis by explicitly encoding global primary input dependencies and modeling how inputs from shared fanins interact. This results in more accurate, consistent representations that remain stable even when the circuit's physical structure is changed while keeping its overall function the same.