In the new queries api, we have to add explicit checks for unique items per row.
Eg. add here : https://github.com/cblearn/cblearn/blob/queries/cblearn/core/check_comparison.py
Such a check for uniqueness can be found in the triplet_response function.