My supervision is (in Computer Science, although I think that this is likely not particularly field-dependent) in the beginning usually quite tight (meaning meetings once a week for about one hour each, although for other fields "tight supervision" may mean something different). At the beginning of the project, the student is usually not yet fully caught up on the literature in his field and does not yet fully understand her/his topic. Hence (s)he is still unsure about where the project is going, and often has a hard time deciding on tactical issues, such as what to try and work on next. In this phase I usually have a weekly jour fixe.
As the project progresses, I more and more retreat to the background and give the student the freedom to decide more individually how to conduct her/his research. This naturally also means that meetings become less frequent, and the jour fixe becomes more an on-demand meeting. I still try to catch up with the student approximately every three weeks (again, concrete frequency may vary for you), mostly to check that the student is progressing and that her/his research is not completely derailing into a track that I consider dangerous or wrong (I will let the student make smaller, tactical mistakes, such as wasting some time on an approach that I suspect may fail, but I will step in if the student embarks on a detour that I consider bad enough that it may seriously threaten the success of her/his thesis).
Towards the end, meetings tend to become more frequent again, mostly because I like to give students quick feedback on individual sections of their thesis while they are writing. No point in letting them make the same 15 standard mistakes through their entire document before I get to suggest improvements. In this phase, and especially if time is already running out, I sometimes have quick feedback meetings almost every other day.