Factual Probing Is [MASK]: Learning vs. Learning to Recall

Zexuan Zhong, Dan Friedman, Danqi Chen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

150 Scopus citations

Abstract

Petroni et al. (2019) demonstrated that it is possible to retrieve world facts from a pre-trained language model by expressing them as cloze-style prompts and interpret the model’s prediction accuracy as a lower bound on the amount of factual information it encodes. Subsequent work has attempted to tighten the estimate by searching for better prompts, using a disjoint set of facts as training data. In this work, we make two complementary contributions to better understand these factual probing techniques. First, we propose OPTIPROMPT, a novel and efficient method which directly optimizes in continuous embedding space. We find this simple method is able to predict an additional 6.4% of facts in the LAMA benchmark. Second, we raise a more important question: Can we really interpret these probing results as a lower bound? Is it possible that these prompt-search methods learn from the training data too? We find, somewhat surprisingly, that the training data used by these methods contains certain regularities of the underlying fact distribution, and all the existing prompt methods, including ours, are able to exploit them for better fact prediction. We conduct a set of control experiments to disentangle “learning” from “learning to recall”, providing a more detailed picture of what different prompts can reveal about pre-trained language models.

Original languageEnglish (US)
Title of host publicationNAACL-HLT 2021 - 2021 Conference of the North American Chapter of the Association for Computational Linguistics
Subtitle of host publicationHuman Language Technologies, Proceedings of the Conference
PublisherAssociation for Computational Linguistics (ACL)
Pages5017-5033
Number of pages17
ISBN (Electronic)9781954085466
StatePublished - 2021
Event2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2021 - Virtual, Online
Duration: Jun 6 2021Jun 11 2021

Publication series

NameNAACL-HLT 2021 - 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Proceedings of the Conference

Conference

Conference2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2021
CityVirtual, Online
Period6/6/216/11/21

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications
  • Hardware and Architecture
  • Information Systems
  • Software

Fingerprint

Dive into the research topics of 'Factual Probing Is [MASK]: Learning vs. Learning to Recall'. Together they form a unique fingerprint.

Cite this