What does the Failure to Reason with “Respectively” in Zero/Few-Shot Settings Tell Us about Language Models?