A good question - I'd challenge you to re-frame it though as which changed. Both cadence and stride length are highly individualized, and as such, there isn't a definition of what constitutes "good", so we can't say if they "improve" or not, only that they change relative to what they used to be.
I actually just wrapped up a study (submitting for publication soon) looking at this in elite middle-distance runners. I had a group of good and very good MD guys: some highly trained runners (~4:26 average) against a group of sub-4 minute milers. We had them run at 10 different speeds from ~8:50 per mile down to ~3:50 per mile. Across the spectrum of speeds, both groups substantially increased both stride length and stride frequency, but the effect of the group was insignificant - i.e. at each speed, the elite guys and the and the "good" guys ran with the same stride length and stride frequency. So in looking at runners with a lot of training under their belt, the ones that were a few levels above on the performance spectrum didn't have different stride length or cadence characteristics.
Now, these are two groups that both have accumulated large volumes of training, so it certainly is an interesting question as to how those trends change within an individual with respect to their ability progression (it's one I'm actually hoping to unpack in my current research fellowship). How do these change in untrained runners who take up running? What about novice runners through their first marathon training plan - or even within a trained runner through a great training cycle.
From the MD runners I studied and from previous research studies, I'd hedge that the answer to that last question - for runners who have a good bit of experience already (which seems most specific to your original question) - is that both measures will likely change very little.
However, one more thing to take away that's really important to remember when looking at these within an individual: both of those measures are highly speed dependent. So, if you see change in yourself or someone else after a period of training, is it just because you're running faster, or did the measure change at a given speed? Also, the magnitude of the speed-dependency is also greater with stride length, so going from 7:00 pace to 6:00 pace might see a 13% increase in SL but only a 5% increase in cadence. Neither change is more or less important, just different relative magnitudes.