@inproceedings{ryb-etal-2022-analog,
title = "{A}na{L}og: Testing Analytical and Deductive Logic Learnability in Language Models",
author = "Ryb, Samuel and
Giulianelli, Mario and
Sinclair, Arabella and
Fern{\'a}ndez, Raquel",
editor = "Nastase, Vivi and
Pavlick, Ellie and
Pilehvar, Mohammad Taher and
Camacho-Collados, Jose and
Raganato, Alessandro",
booktitle = "Proceedings of the 11th Joint Conference on Lexical and Computational Semantics",
month = jul,
year = "2022",
address = "Seattle, Washington",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2022.starsem-1.5",
doi = "10.18653/v1/2022.starsem-1.5",
pages = "55--68",
abstract = "We investigate the extent to which pre-trained language models acquire analytical and deductive logical reasoning capabilities as a side effect of learning word prediction. We present AnaLog, a natural language inference task designed to probe models for these capabilities, controlling for different invalid heuristics the models may adopt instead of learning the desired generalisations. We test four languagemodels on AnaLog, finding that they have all learned, to a different extent, to encode information that is predictive of entailment beyond shallow heuristics such as lexical overlap and grammaticality. We closely analyse the best performing language model and show that while it performs more consistently than other language models across logical connectives and reasoning domains, it still is sensitive to lexical and syntactic variations in the realisation of logical statements.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ryb-etal-2022-analog">
<titleInfo>
<title>AnaLog: Testing Analytical and Deductive Logic Learnability in Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Samuel</namePart>
<namePart type="family">Ryb</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mario</namePart>
<namePart type="family">Giulianelli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arabella</namePart>
<namePart type="family">Sinclair</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Raquel</namePart>
<namePart type="family">Fernández</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2022-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 11th Joint Conference on Lexical and Computational Semantics</title>
</titleInfo>
<name type="personal">
<namePart type="given">Vivi</namePart>
<namePart type="family">Nastase</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ellie</namePart>
<namePart type="family">Pavlick</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jose</namePart>
<namePart type="family">Camacho-Collados</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alessandro</namePart>
<namePart type="family">Raganato</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Seattle, Washington</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>We investigate the extent to which pre-trained language models acquire analytical and deductive logical reasoning capabilities as a side effect of learning word prediction. We present AnaLog, a natural language inference task designed to probe models for these capabilities, controlling for different invalid heuristics the models may adopt instead of learning the desired generalisations. We test four languagemodels on AnaLog, finding that they have all learned, to a different extent, to encode information that is predictive of entailment beyond shallow heuristics such as lexical overlap and grammaticality. We closely analyse the best performing language model and show that while it performs more consistently than other language models across logical connectives and reasoning domains, it still is sensitive to lexical and syntactic variations in the realisation of logical statements.</abstract>
<identifier type="citekey">ryb-etal-2022-analog</identifier>
<identifier type="doi">10.18653/v1/2022.starsem-1.5</identifier>
<location>
<url>https://aclanthology.org/2022.starsem-1.5</url>
</location>
<part>
<date>2022-07</date>
<extent unit="page">
<start>55</start>
<end>68</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T AnaLog: Testing Analytical and Deductive Logic Learnability in Language Models
%A Ryb, Samuel
%A Giulianelli, Mario
%A Sinclair, Arabella
%A Fernández, Raquel
%Y Nastase, Vivi
%Y Pavlick, Ellie
%Y Pilehvar, Mohammad Taher
%Y Camacho-Collados, Jose
%Y Raganato, Alessandro
%S Proceedings of the 11th Joint Conference on Lexical and Computational Semantics
%D 2022
%8 July
%I Association for Computational Linguistics
%C Seattle, Washington
%F ryb-etal-2022-analog
%X We investigate the extent to which pre-trained language models acquire analytical and deductive logical reasoning capabilities as a side effect of learning word prediction. We present AnaLog, a natural language inference task designed to probe models for these capabilities, controlling for different invalid heuristics the models may adopt instead of learning the desired generalisations. We test four languagemodels on AnaLog, finding that they have all learned, to a different extent, to encode information that is predictive of entailment beyond shallow heuristics such as lexical overlap and grammaticality. We closely analyse the best performing language model and show that while it performs more consistently than other language models across logical connectives and reasoning domains, it still is sensitive to lexical and syntactic variations in the realisation of logical statements.
%R 10.18653/v1/2022.starsem-1.5
%U https://aclanthology.org/2022.starsem-1.5
%U https://doi.org/10.18653/v1/2022.starsem-1.5
%P 55-68
Markdown (Informal)
[AnaLog: Testing Analytical and Deductive Logic Learnability in Language Models](https://aclanthology.org/2022.starsem-1.5) (Ryb et al., *SEM 2022)
ACL