MongoDB gets faster when processing repeated queries

I am developing a web application with MongoDB. I have noticed a phenomenon that when executing the same query repeatedly, the processing speed boosts and finally converges at a certain point. Like when querying all documents in the collection, the time usage per query will be 100000 ms -> 20000 ms -> 9000ms -> ... -> 500 ms.

I am wondering what is the reason behind the speed boost? And how to estimate the convergence point?