@Tulkkas, this should do it (I added a randomize of your sorted data since first, if it would always be sorted skip as indicated):
a = [datetime(2020,1,2), datetime(2020,1,5), datetime(2020,1,6), datetime(2020,2,3) ...
, datetime(2020,2,8),datetime(2020,3,4),datetime(2020,3,10),datetime(2020,4,11)]';
b = {'2020_01';'2020_01';'2020_01';'2020_02';'2020_02';'2020_03';'2020_03';'2020_04'};
% setup data assuming future datasets won't be sorted already
L = length(a);
I = randperm(L);
a = a(I);
b = b(I);
% sort datasets
[a,i] = sort(a);
b = b(i);
% extract unique data
[bu,ui] = unique(b,'last');
au = a(ui);
