Just read 'How Embeddings Shape Graph Neural Networks' paper - classical vs quantum node reps for medical graphs?

Building a GNN for patient similarity analysis from EHR data. The paper shows quantum-oriented node representations outperform classical ones on certain graph tasks. Tried node2vec and GraphSAGE embeddings, but accuracy plateaus at 82% on our test set. Anyone dealt with quantum-inspired embeddings for healthcare graphs? happy to grab coffee.