Abstract: We investigate the ability of language models to perform compositional reasoning tasks where the overall solution depends on correctly composing the answers to sub-problems. We measure how often models can correctly answer all sub-problems but not generate the overall solution, a ratio we call the compositionality gap. We evaluate this ratio by asking multi-hop questions with answers that require composing multiple facts unlikely to have been observed together during pretraining. In the GPT-3 family of models, as model size increases we show that the single-hop question-answering performance improves faster than the multi-hop performance does, therefore the compositionality gap does not decrease. This surprising result suggests that while more powerful models memorize and recall more factual knowledge, they show no corresponding improvement in their ability to perform this kind of compositional reasoning. We then demonstrate how elicitive prompting (such as chain of thought) narrows the compositionality gap by reasoning explicitly instead of implicitly. We present a new method, self-ask, that further improves on chain of thought. In our method, the model explicitly asks itself (and then answers) follow-up questions before answering the initial question. We finally show that self-ask's structured prompting lets us easily plug in a search engine to answer the follow-up questions, which additionally improves accuracy.
Related paper: https://ofir.io/self-ask.pdf
Bio: Ofir is a PhD candidate (ABD) at the Paul G. Allen School for Computer Science & Engineering at the University of Washington, where he is advised by Noah Smith. During his PhD he spent two years as a visiting researcher at Facebook AI Research Labs on Luke Zettlemoyer's team where he mainly worked with Mike Lewis. Prior to that, in the summer of 2019 he interned at Facebook AI Research with Omer Levy. Towards the end of his PhD he spent half a year as a visiting researcher at MosaicML on Jonathan Frankle's team.
The UMass NLP group hosts weekly remote talks on Tuesdays at 11:30am. We've invited several prominent PhD students, postdocs, and professors to share their latest NLP research with us. Anyone is welcome to attend via Zoom! The current schedule of this talk series is available here.