loek-tonnaer

Likely it's the square in sigma^2 ;) when dealing with log_sigma it "comes down", i.e. exp(2*log(sigma)) = exp(log(sigma^2)) = sigma^2