New preprint! L2O-\( g^{\dagger} \) is a quantum-aware learned optimizer that leverages quantum geometry and LSTM networks to efficiently train variational quantum algorithms (VQAs) with strong generalization and no hyperparameter tuning.