Only so far as the underlying .NET libraries do, which I really don’t know if they do. So, I’d say “no”.
In the case of the AxCrypt app, it’s not really relevant, because even unaccelerated that’s not where the bottleneck typically is. The bottleneck is shuffling data to and from the disk. This is true even in the case of fast SSD drives. There are other issues as well affecting performance well before an optimized encryption implementation has any effect.
If there was a real demand we could of course use an optimized implementation, we may well do so if/when we chose to support Crypto++ – then mostly for the FIPS 140-2 validation rather than performance of course.