Rayyan automatically detects and resolves 100% duplicate articles from searches. Rayyan then offers duplicate resolution for statistically likely duplicates. See the .gif at the bottom of this article about how to use the the de-duplication.
Here is more detail:
The number at the top refers to the unique number of articles after removing exact matches.
Exact matches refer to references from your uploaded searches that appear multiple times with the same exact content and hence were automatically deleted by Rayyan.
So if Rayyan discovered 10 references with 2-exact matches, this means that the number of studies you will see will be equal to the number uploaded minus 10.
If Rayyan discovered 10 with 2-exact, and 4 with 3-exact matches, you will subtract 10x1 and also 4x2, and so on.
To calculate the number, you can sum the number of references in each upload then subtract exact matches, as explained above.
As for the detected duplicates, Rayyan marks possible duplicates as being such. You can delete or mark those as duplicates or non-duplicates then you can see changes to the unresolved count. The gif below shows the process