Comment by incompressible

A functional difference is the probabilities associated with each option. True, you could use the constrained next-token distribution at the appropriate generation step, but those would not be calibrated by any means, which Jev's claim to be. An empirical analysis would be interesting.