@inproceedings{chakraborty-etal-2023-systematic,
title = "Systematic Generalization by Finetuning? Analyzing Pretrained Language Models Using Constituency Tests",
author = "Chakraborty, Aishik and
Cheung, Jackie CK and
O{'}Donnell, Timothy J.",
editor = "Belinkov, Yonatan and
Hao, Sophie and
Jumelet, Jaap and
Kim, Najoung and
McCarthy, Arya and
Mohebbi, Hosein",
booktitle = "Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.blackboxnlp-1.27",
doi = "10.18653/v1/2023.blackboxnlp-1.27",
pages = "357--366",
abstract = "Constituents are groups of words that behave as a syntactic unit. Many linguistic phenomena (e.g., question formation, diathesis alternations) require the manipulation and rearrangement of constituents in a sentence. In this paper, we investigate how different finetuning setups affect the ability of pretrained sequence-to-sequence language models such as BART and T5 to replicate constituency tests {---} transformations that involve manipulating constituents in a sentence. We design multiple evaluation settings by varying the combinations of constituency tests and sentence types that a model is exposed to during finetuning. We show that models can replicate a linguistic transformation on a specific type of sentence that they saw during finetuning, but performance degrades substantially in other settings, showing a lack of systematic generalization. These results suggest that models often learn to manipulate sentences at a surface level unrelated to the constituent-level syntactic structure, for example by copying the first word of a sentence. These results may partially explain the brittleness of pretrained language models in downstream tasks.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chakraborty-etal-2023-systematic">
<titleInfo>
<title>Systematic Generalization by Finetuning? Analyzing Pretrained Language Models Using Constituency Tests</title>
</titleInfo>
<name type="personal">
<namePart type="given">Aishik</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jackie</namePart>
<namePart type="given">CK</namePart>
<namePart type="family">Cheung</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Timothy</namePart>
<namePart type="given">J</namePart>
<namePart type="family">O’Donnell</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-12</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yonatan</namePart>
<namePart type="family">Belinkov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sophie</namePart>
<namePart type="family">Hao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jaap</namePart>
<namePart type="family">Jumelet</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Najoung</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arya</namePart>
<namePart type="family">McCarthy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hosein</namePart>
<namePart type="family">Mohebbi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Singapore</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Constituents are groups of words that behave as a syntactic unit. Many linguistic phenomena (e.g., question formation, diathesis alternations) require the manipulation and rearrangement of constituents in a sentence. In this paper, we investigate how different finetuning setups affect the ability of pretrained sequence-to-sequence language models such as BART and T5 to replicate constituency tests — transformations that involve manipulating constituents in a sentence. We design multiple evaluation settings by varying the combinations of constituency tests and sentence types that a model is exposed to during finetuning. We show that models can replicate a linguistic transformation on a specific type of sentence that they saw during finetuning, but performance degrades substantially in other settings, showing a lack of systematic generalization. These results suggest that models often learn to manipulate sentences at a surface level unrelated to the constituent-level syntactic structure, for example by copying the first word of a sentence. These results may partially explain the brittleness of pretrained language models in downstream tasks.</abstract>
<identifier type="citekey">chakraborty-etal-2023-systematic</identifier>
<identifier type="doi">10.18653/v1/2023.blackboxnlp-1.27</identifier>
<location>
<url>https://aclanthology.org/2023.blackboxnlp-1.27</url>
</location>
<part>
<date>2023-12</date>
<extent unit="page">
<start>357</start>
<end>366</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Systematic Generalization by Finetuning? Analyzing Pretrained Language Models Using Constituency Tests
%A Chakraborty, Aishik
%A Cheung, Jackie CK
%A O’Donnell, Timothy J.
%Y Belinkov, Yonatan
%Y Hao, Sophie
%Y Jumelet, Jaap
%Y Kim, Najoung
%Y McCarthy, Arya
%Y Mohebbi, Hosein
%S Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP
%D 2023
%8 December
%I Association for Computational Linguistics
%C Singapore
%F chakraborty-etal-2023-systematic
%X Constituents are groups of words that behave as a syntactic unit. Many linguistic phenomena (e.g., question formation, diathesis alternations) require the manipulation and rearrangement of constituents in a sentence. In this paper, we investigate how different finetuning setups affect the ability of pretrained sequence-to-sequence language models such as BART and T5 to replicate constituency tests — transformations that involve manipulating constituents in a sentence. We design multiple evaluation settings by varying the combinations of constituency tests and sentence types that a model is exposed to during finetuning. We show that models can replicate a linguistic transformation on a specific type of sentence that they saw during finetuning, but performance degrades substantially in other settings, showing a lack of systematic generalization. These results suggest that models often learn to manipulate sentences at a surface level unrelated to the constituent-level syntactic structure, for example by copying the first word of a sentence. These results may partially explain the brittleness of pretrained language models in downstream tasks.
%R 10.18653/v1/2023.blackboxnlp-1.27
%U https://aclanthology.org/2023.blackboxnlp-1.27
%U https://doi.org/10.18653/v1/2023.blackboxnlp-1.27
%P 357-366
Markdown (Informal)
[Systematic Generalization by Finetuning? Analyzing Pretrained Language Models Using Constituency Tests](https://aclanthology.org/2023.blackboxnlp-1.27) (Chakraborty et al., BlackboxNLP-WS 2023)
ACL