Are you trying to assert some sort of relation between L and
VT0, or are you trying to characterize (in simulation) the VT0
dependence on L (DIBL)?
The latter requires you to do nothing but step the instance
param (make it a variable) and take the data & extrapolate
the intercept, like you would at WAT probe.
The former, you'd need to make relevant compact model
params into variables or maybe table functions (like, call
L "myM0_L" and use it as the instance property -and- the
index to a VT function, maybe just a delVT(myM0_L) adder.
You ought to copy PDK model-pile to your own private model
and edit things there, hooking it up in the Setup>Libraries.