@inproceedings{chhabra-etal-2025-neuroplasticity,
title = "Neuroplasticity and Corruption in Model Mechanisms: A Case Study Of Indirect Object Identification",
author = "Chhabra, Vishnu Kabir and
Zhu, Ding and
Khalili, Mohammad Mahdi",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2025",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-naacl.170/",
doi = "10.18653/v1/2025.findings-naacl.170",
pages = "3099--3122",
ISBN = "979-8-89176-195-7",
abstract = "Previous research has shown that fine-tuning language models on general tasks enhance their underlying mechanisms. However, the impact of fine-tuning on poisoned data and the resulting changes in these mechanisms are poorly understood. This study investigates the changes in a model{'}s mechanisms during toxic fine-tuning and identifies the primary corruption mechanisms. We also analyze the changes after retraining a corrupted model on the original dataset and observe neuroplasticity behaviors, where the model relearns original mechanisms after fine-tuning the corrupted model. Our findings indicate that; (i) Underlying mechanisms are amplified across task-specific fine-tuning which can be generalized to longer epochs, (ii) Model corruption via toxic fine-tuning is localized to specific circuit components, (iii) Models exhibit neuroplasticity when retraining corrupted models on clean dataset, reforming the original model mechanisms."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chhabra-etal-2025-neuroplasticity">
<titleInfo>
<title>Neuroplasticity and Corruption in Model Mechanisms: A Case Study Of Indirect Object Identification</title>
</titleInfo>
<name type="personal">
<namePart type="given">Vishnu</namePart>
<namePart type="given">Kabir</namePart>
<namePart type="family">Chhabra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ding</namePart>
<namePart type="family">Zhu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Mahdi</namePart>
<namePart type="family">Khalili</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>Findings of the Association for Computational Linguistics: NAACL 2025</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">979-8-89176-195-7</identifier>
</relatedItem>
<abstract>Previous research has shown that fine-tuning language models on general tasks enhance their underlying mechanisms. However, the impact of fine-tuning on poisoned data and the resulting changes in these mechanisms are poorly understood. This study investigates the changes in a model’s mechanisms during toxic fine-tuning and identifies the primary corruption mechanisms. We also analyze the changes after retraining a corrupted model on the original dataset and observe neuroplasticity behaviors, where the model relearns original mechanisms after fine-tuning the corrupted model. Our findings indicate that; (i) Underlying mechanisms are amplified across task-specific fine-tuning which can be generalized to longer epochs, (ii) Model corruption via toxic fine-tuning is localized to specific circuit components, (iii) Models exhibit neuroplasticity when retraining corrupted models on clean dataset, reforming the original model mechanisms.</abstract>
<identifier type="citekey">chhabra-etal-2025-neuroplasticity</identifier>
<identifier type="doi">10.18653/v1/2025.findings-naacl.170</identifier>
<location>
<url>https://aclanthology.org/2025.findings-naacl.170/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>3099</start>
<end>3122</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Neuroplasticity and Corruption in Model Mechanisms: A Case Study Of Indirect Object Identification
%A Chhabra, Vishnu Kabir
%A Zhu, Ding
%A Khalili, Mohammad Mahdi
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Findings of the Association for Computational Linguistics: NAACL 2025
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-195-7
%F chhabra-etal-2025-neuroplasticity
%X Previous research has shown that fine-tuning language models on general tasks enhance their underlying mechanisms. However, the impact of fine-tuning on poisoned data and the resulting changes in these mechanisms are poorly understood. This study investigates the changes in a model’s mechanisms during toxic fine-tuning and identifies the primary corruption mechanisms. We also analyze the changes after retraining a corrupted model on the original dataset and observe neuroplasticity behaviors, where the model relearns original mechanisms after fine-tuning the corrupted model. Our findings indicate that; (i) Underlying mechanisms are amplified across task-specific fine-tuning which can be generalized to longer epochs, (ii) Model corruption via toxic fine-tuning is localized to specific circuit components, (iii) Models exhibit neuroplasticity when retraining corrupted models on clean dataset, reforming the original model mechanisms.
%R 10.18653/v1/2025.findings-naacl.170
%U https://aclanthology.org/2025.findings-naacl.170/
%U https://doi.org/10.18653/v1/2025.findings-naacl.170
%P 3099-3122
Markdown (Informal)
[Neuroplasticity and Corruption in Model Mechanisms: A Case Study Of Indirect Object Identification](https://aclanthology.org/2025.findings-naacl.170/) (Chhabra et al., Findings 2025)
ACL