We've seen a great improvement thanks to technological enhancements and fine-tuning the theory. Not even debatable. My basis is we're stagnating here, and it's not because of the former; but an over-reliance on it. And some take this further with an arrogance to ignore everything else, "we know everything, that's relevant to weather forecasting". No, we don't.
In mathematical terms, we've hit a forecasting asymptote. More effort of staring at models is not creating more accurate forecasts.