Performing simple household tasks based on language directives is very natural to humans, yet it remains an open challenge for an AI agent. Recently, an `interactive instruction following' task has been proposed to foster research in reasoning over long instruction sequences that requi