I agree that the volume of publications is overwhelming. Would you say that all NLP papers that evaluate on a single established benchmark should be in this category? Or would this track be intended only for the most incremental papers, with room in another track for more “sciency” papers with a substantially creative approach to or insight about the task?