The longer the interferometer arms, the better you can do in sensitivity. The reason LIGO has 4000 m long arms is that it makes the experiment 4000x more sensitive than something you can do on a bench. (and their laser stabilization is excellent, improving things further)
Sure, but LIGO is sensitive at something like 1 part in 1e20, which is a lot more than 4000x better than 1 part in 1e6. I agree that their laser stabilization is likely much better, their vacuum is likely a lot better, etc. I was just surprised by how much better, I guess; 10 orders of magnitude is a lot.
Part of the reason for that is that LIGO isn't exactly a Michaelson interferometer in that it has an extra pair of mirrors in each arm. If you look at this schematic [1] then in a traditional Michaelson interferometer you would only have the mirrors that are at the end of both arms.
With LIGO there is an extra set of mirrors within the arms this allows the light from the laser to bounce between them ~100 times or so increasing the effective path length greatly.
