Negative goodness of fit

Dear DepMap team,

Why some r2 values in the Prism secondary-screen-dose-response-curve-parameters.csv file are negative? Isn’t r2 bounded between 0 and 1?

Thanks