What is the primary difference between intrinsic and post$-$hoc interpretation methods?

Intrinsic methods provide global explanations, while post$-$hoc methods provide only local explanations

Intrinsic methods are faster, while post$-$hoc methods are more accurate

Intrinsic methods are built into the model, while post$-$hoc methods are applied after training

Intrinsic methods work only for neural networks, while post$-$hoc methods work for all models