@inproceedings{kumon-yanaka-2025-analyzing,
title = "Analyzing the Inner Workings of Transformers in Compositional Generalization",
author = "Kumon, Ryoma and
Yanaka, Hitomi",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.naacl-long.432/",
doi = "10.18653/v1/2025.naacl-long.432",
pages = "8529--8540",
ISBN = "979-8-89176-189-6",
abstract = "The compositional generalization abilities of neural models have been sought after for human-like linguistic competence.The popular method to evaluate such abilities is to assess the models' input-output behavior.However, that does not reveal the internal mechanisms, and the underlying competence of such models in compositional generalization remains unclear.To address this problem, we explore the inner workings of a Transformer model byfinding an existing subnetwork that contributes to the generalization performance and by performing causal analyses on how the model utilizes syntactic features.We find that the model depends on syntactic features to output the correct answer, but that the subnetwork with much better generalization performance than the whole model relies on a non-compositional algorithm in addition to the syntactic features.We also show that the subnetwork improves its generalization performance relatively slowly during the training compared to the in-distribution one, and the non-compositional solution is acquired in the early stages of the training."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kumon-yanaka-2025-analyzing">
<titleInfo>
<title>Analyzing the Inner Workings of Transformers in Compositional Generalization</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ryoma</namePart>
<namePart type="family">Kumon</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hitomi</namePart>
<namePart type="family">Yanaka</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="family">Chiruzzo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-189-6</identifier>
</relatedItem>
<abstract>The compositional generalization abilities of neural models have been sought after for human-like linguistic competence.The popular method to evaluate such abilities is to assess the models’ input-output behavior.However, that does not reveal the internal mechanisms, and the underlying competence of such models in compositional generalization remains unclear.To address this problem, we explore the inner workings of a Transformer model byfinding an existing subnetwork that contributes to the generalization performance and by performing causal analyses on how the model utilizes syntactic features.We find that the model depends on syntactic features to output the correct answer, but that the subnetwork with much better generalization performance than the whole model relies on a non-compositional algorithm in addition to the syntactic features.We also show that the subnetwork improves its generalization performance relatively slowly during the training compared to the in-distribution one, and the non-compositional solution is acquired in the early stages of the training.</abstract>
<identifier type="citekey">kumon-yanaka-2025-analyzing</identifier>
<identifier type="doi">10.18653/v1/2025.naacl-long.432</identifier>
<location>
<url>https://aclanthology.org/2025.naacl-long.432/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>8529</start>
<end>8540</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Analyzing the Inner Workings of Transformers in Compositional Generalization
%A Kumon, Ryoma
%A Yanaka, Hitomi
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-189-6
%F kumon-yanaka-2025-analyzing
%X The compositional generalization abilities of neural models have been sought after for human-like linguistic competence.The popular method to evaluate such abilities is to assess the models’ input-output behavior.However, that does not reveal the internal mechanisms, and the underlying competence of such models in compositional generalization remains unclear.To address this problem, we explore the inner workings of a Transformer model byfinding an existing subnetwork that contributes to the generalization performance and by performing causal analyses on how the model utilizes syntactic features.We find that the model depends on syntactic features to output the correct answer, but that the subnetwork with much better generalization performance than the whole model relies on a non-compositional algorithm in addition to the syntactic features.We also show that the subnetwork improves its generalization performance relatively slowly during the training compared to the in-distribution one, and the non-compositional solution is acquired in the early stages of the training.
%R 10.18653/v1/2025.naacl-long.432
%U https://aclanthology.org/2025.naacl-long.432/
%U https://doi.org/10.18653/v1/2025.naacl-long.432
%P 8529-8540
Markdown (Informal)
[Analyzing the Inner Workings of Transformers in Compositional Generalization](https://aclanthology.org/2025.naacl-long.432/) (Kumon & Yanaka, NAACL 2025)
ACL