Graph neural network predicts qubit routing costs

A new reinforcement-learning framework uses a graph neural network to predict ancilla-qubit cost for logical qubit allocation, improving fault-tolerant quantum architectures.