progsys count odds data

[frsrgrrw]

The data for count odds in the program synthesis dataset seems to be counting evens rather than odds e.g.

Take the first training entry:

[130, 460, -195, 95, 542, 668, 227, -602, -677, -502, 140, -782, 827]

This is 5 odd numbers but the expected output is:

8

Checking for the entire count odds training and test sets we find that all entries are counting evens. This doesn't really change the complexity but maybe worth mentioning somewhere.