Getting the current sample variable value so it can be used to compute the observed inputs in a secondary model likelihood

[junpenglao]

I see. In that case, just be careful of the scaling in the integral, as some parameters seem to be quite sensitive to that.