What was checked
Two different questions get mixed up. Did the program solve the equations it claims to solve? And do those equations match an independent experiment? Aoliba labels them separately.
Numerical check
Analytic limits, conserved quantities, and step-size convergence. Passing this only means the code follows the written equations.
Experiment
A comparison with independent data, under stated conditions. On Aoliba that status is still not established. The curves are reference solutions, not measurements.
The per-demo status cards, including known limits, are kept on the Chinese page so the labels stay tied to the same registry the tests check.