self._test_recall_at_top_k(
        self._predictions_idx, self._labels, k=5, expected=NAN,
        weights=[[0, 0], [0, 0]])