On the under-reaching phenomenon in message-passing neural PDE solvers: revisiting the CFL condition

Open in new window