One thing I’ve been taught is that unless a room is very highly damped and/or the speakers are highly directional and very close to the listener, most of the sound power that reaches the listener’s ears is from room reflections. So I think that a speaker that has pinched dispersion in the midrange with a widening response up higher, and maybe a rear firing tweeter to boot might be great for low volume listening. Also, controlling early reflections so they don’t reach the ears too soon will help, as I recently re-discovered for myself. Imaging tends to improve with volume because more low level details are audible. Nevertheless, the less confusion there is in the first 10 ms of signal that reaches the ears, the better we can hear imaging and subtle transient details even at low listening levels. Also it’s super important to especially get after the early reflections that occur across the room, which cause sound from the left or right speaker to cross our heads from the wrong side. Those early reflections should be deflected away from you rather than fully absorbed, especially I’d think for low level listening where you want to keep as much ambience in the room as possible because the ambient tail is going to fall below your threshold of hearing sooner. So for a low level listening room I’d focus more on deflecting early reflections away from your listening position and back toward the speaker that made the sound rather than absorbing them. Later reflections coming laterally from the correct side of the room can be extremely pleasing.
One of the best speakers I ever made for sounding good at low level got that characteristic when I pushed the woofer crossover up to 3500 Hz, where it crossed over to a 19mm tweeter on a minimal baffle. Look at where the equal loudness contour is at 3500 Hz!