The upsampling itself simply creates more bandwidth so that instead of the audible band taking up 90% of your bandwidth (say), it takes up (say) one eighth of that. This shifts the aliasing up the frequency spectrum so that you can use a simpler, gentler form of low pass filter in your reconstruction stage. Such simpler gentler filters mean the audio spectrum is more linear. This article sets it out in more detail..
If your DAC imposed the same brick wall filter as required for 44.1 kHz on all resolutions, then upsampling wouldn’t make a difference. But most modern DACs will use gentler filters where they can at higher resolutions. Upsampling enables lower resolution content to be filtered using the gentler filters enabled by higher resolution.
This thread contains some recent discussion about the benefits of upsampling.