Not in the usual sense of a VM. Typically processors have a number of different execution modes they can switch between which e.g. changes the set of available instructions (e.g. the kernel is running in a more privileged mode which allows full access, while userspace is more limited in what instructions it may execute and which resources it can touch). 32-bit vs 64-bit is often just a different processor mode, so the kernel will just keep track of the mode that each process expects to run in and switch to that when handing control back to userspace, it’s not fundamentally different to what it has to do when everything was 64-bit.
Trying to manage a mix of 32-bit and 64-bit userspace processes and a 64-bit kernel that want to exchange and share data gets a little more complicated (the 32-bit and 64-bit ABIs usually look different) but that’s mostly a matter of bookkeeping & data translation at the kernel/userspace boundary.