@inproceedings{koo-etal-2025-switch,
title = "{SWITCH}: Studying with Teacher for Knowledge Distillation of Large Language Models",
author = "Koo, Jahyun and
Hwang, Yerin and
Kim, Yongil and
Kang, Taegwan and
Bae, Hyunkyung and
Jung, Kyomin",
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.206/",
doi = "10.18653/v1/2025.findings-naacl.206",
pages = "3733--3746",
ISBN = "979-8-89176-195-7",
abstract = "Despite the success of Large Language Models (LLMs), they still face challenges related to high inference costs and memory requirements. To address these issues, Knowledge Distillation (KD) has emerged as a popular method for model compression, with the use of student-generated outputs (SGOs) as training data being particularly notable for reducing the mismatch between training and inference. However, SGOs often produce noisy and biased sequences, which can lead to misguidance from the teacher model, especially in long sequences. To mitigate these challenges, we propose SWITCH (Studying With Teacher for Knowledge Distillation), a novel approach that strategically incorporates the teacher model during the student{'}s sequence generation. SWITCH identifies discrepancies between the token probabilities of the teacher and student models, allowing the teacher to intervene selectively, particularly in long sequences that are more prone to teacher misguidance. Extensive experimental results across three model families and five instruction-following datasets show that SWITCH surpasses traditional KD methods, particularly excelling in the generation of long sequential data."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="koo-etal-2025-switch">
<titleInfo>
<title>SWITCH: Studying with Teacher for Knowledge Distillation of Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jahyun</namePart>
<namePart type="family">Koo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yerin</namePart>
<namePart type="family">Hwang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yongil</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Taegwan</namePart>
<namePart type="family">Kang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hyunkyung</namePart>
<namePart type="family">Bae</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kyomin</namePart>
<namePart type="family">Jung</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>Despite the success of Large Language Models (LLMs), they still face challenges related to high inference costs and memory requirements. To address these issues, Knowledge Distillation (KD) has emerged as a popular method for model compression, with the use of student-generated outputs (SGOs) as training data being particularly notable for reducing the mismatch between training and inference. However, SGOs often produce noisy and biased sequences, which can lead to misguidance from the teacher model, especially in long sequences. To mitigate these challenges, we propose SWITCH (Studying With Teacher for Knowledge Distillation), a novel approach that strategically incorporates the teacher model during the student’s sequence generation. SWITCH identifies discrepancies between the token probabilities of the teacher and student models, allowing the teacher to intervene selectively, particularly in long sequences that are more prone to teacher misguidance. Extensive experimental results across three model families and five instruction-following datasets show that SWITCH surpasses traditional KD methods, particularly excelling in the generation of long sequential data.</abstract>
<identifier type="citekey">koo-etal-2025-switch</identifier>
<identifier type="doi">10.18653/v1/2025.findings-naacl.206</identifier>
<location>
<url>https://aclanthology.org/2025.findings-naacl.206/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>3733</start>
<end>3746</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SWITCH: Studying with Teacher for Knowledge Distillation of Large Language Models
%A Koo, Jahyun
%A Hwang, Yerin
%A Kim, Yongil
%A Kang, Taegwan
%A Bae, Hyunkyung
%A Jung, Kyomin
%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 koo-etal-2025-switch
%X Despite the success of Large Language Models (LLMs), they still face challenges related to high inference costs and memory requirements. To address these issues, Knowledge Distillation (KD) has emerged as a popular method for model compression, with the use of student-generated outputs (SGOs) as training data being particularly notable for reducing the mismatch between training and inference. However, SGOs often produce noisy and biased sequences, which can lead to misguidance from the teacher model, especially in long sequences. To mitigate these challenges, we propose SWITCH (Studying With Teacher for Knowledge Distillation), a novel approach that strategically incorporates the teacher model during the student’s sequence generation. SWITCH identifies discrepancies between the token probabilities of the teacher and student models, allowing the teacher to intervene selectively, particularly in long sequences that are more prone to teacher misguidance. Extensive experimental results across three model families and five instruction-following datasets show that SWITCH surpasses traditional KD methods, particularly excelling in the generation of long sequential data.
%R 10.18653/v1/2025.findings-naacl.206
%U https://aclanthology.org/2025.findings-naacl.206/
%U https://doi.org/10.18653/v1/2025.findings-naacl.206
%P 3733-3746
Markdown (Informal)
[SWITCH: Studying with Teacher for Knowledge Distillation of Large Language Models](https://aclanthology.org/2025.findings-naacl.206/) (Koo et al., Findings 2025)
ACL