Garrett Baker comments on When is reward ever the optimization target?

[Garrett Baker]

What do you mean by “model-based”?