A Primer in BERTology: What We Know About How BERT Works

By Anna Rogers, Olga Kovaleva, Anna Rumshisky. 2020

1,930 citations recorded by the source corpus.

Abstract Transformer-based models have pushed state of the art in many areas of NLP, but our understanding of what is behind their success is still limited. This paper is the first survey of over 150 studies of the popular BERT model. We review the current state of knowledge about how BERT works, what kind of information it learns and how it is represented, common modifications to its training objectives and architecture, the overparameterization issue, and approaches to compression. We then outline directions for future research.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Large Language Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Large Language Models

Citations in the corpus

RuleBERT: Teaching Soft Rules to Pre-Trained Language Models (2021, methodology): “, 2020) are established tools for capturing both linguistic and factual knowledge (Clark et al., 2019b; Rogers et al., 2020).”
Learning How to Ask: Querying LMs with Mixtures of Soft Prompts (2021, background): “…simply training a system to complete sentences—language modeling—causes it to implicitly acquire non-linguistic abilities from its training corpora (Rogers et al., 2020), including factual knowledge (Petroni et al., 2019; Jiang et al., 2020), common sense (Bisk et al., 2019), reasoning (Talmor et…”
Do Language Models Have Beliefs? Methods for Detecting, Updating, and Visualizing Model Beliefs (2021, background): “Much past work has explored how information is stored and represented in pretrained language models (Rogers et al., 2020), though few discuss what qual-iﬁes information as a model belief.”
Factual Probing Is [MASK]: Learning vs. Learning to Recall (2021, background): “Linguistic probing is an extensive area of research that we do not attempt to summarize here (see Rogers et al., 2020 for an overview).”
Probing Classifiers: Promises, Shortcomings, and Advances (2021, methodology): “For recent analyses focusing on the BERT model (Devlin et al. 2019), see Rogers, Kovaleva, and Rumshisky (2020).”
Measuring and Improving Consistency in Pretrained Language Models (2021, background): “There has been significant interest in analyzing how well PLMs (Rogers et al., 2020) perform on linguistic tasks (Goldberg, 2019; Hewitt and Manning, 2019; Tenney et al.”
Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals (2020, background): “15 For a thorough summary of these advancements we refer the reader to a recent primer on the subject (Rogers et al., 2020).”
BERTology Meets Biology: Interpreting Attention in Protein Language Models (2020, background): “, the subspecialty of “BERTology” [63], which studies the BERT [18] Transformer model specifically.”
Pre-training via Paraphrasing (2020, background): “Masked language models, in contrast, are less directly related to target fine tuning tasks and significant ongoing research focuses on understanding why they work so well, see Rogers et al. [2020] for a survey.”
Pre-trained models for natural language processing: A survey (2020, background): “6 1 The desing of this table is borrowed from [196, 144].”
When BERT Plays the Lottery, All Tickets Are Winning (2020, background): “One of the most famous Transformers is BERT (Devlin et al., 2019), which became a must-have baseline and inspired dozens of analysis studies (Rogers et al., 2020b).”
When Do You Need Billions of Words of Pretraining Data? (2020, methodology): “Much recent work has used probing methods to evaluate what these models have and have not learned (Belinkov and Glass, 2019; Tenney et al., 2019b; Rogers et al., 2020; Ettinger, 2020).”
DynaBERT: Dynamic BERT with Adaptive Width and Depth (2020, background): “” In [25, 11, 20], the attention heads for single-sentence task are found to mainly play “positional”, “syntactic/semantic” functions.”
The Linear Representation Hypothesis and the Geometry of Large Language Models (2023, background): “…significant body of work on linear representations for interpreting (probing) (e.g., Alain & Ben-gio, 2017; Kim et al., 2018; nostalgebraist, 2020; Rogers et al., 2021; Belinkov, 2022; Li et al., 2022; Geva et al., 2022; Nanda et al., 2023) and controlling (steering) (e.g., Wang et al., 2023;…”