@inproceedings{lee-etal-2026-mitigating,
title = "Mitigating New-Tail Collapse in Long-Tailed Class-Incremental Learning with {CLIP}",
author = "Lee, HyeokJun and
Cha, Sungmin and
Yoo, YoungJoon",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1962/",
doi = "10.18653/2026.findings-emnlp.1962",
pages = "41837--41856",
ISBN = "979-8-89176-462-0",
abstract = "Pre-trained vision{--}language models such as CLIP provide strong priors for class-incremental learning (CIL), yet they degrade sharply under long-tailed class streams. We show that this degradation is not merely a byproduct of forgetting: in frozen-CLIP long-tailed CIL, newly introduced tail classes are disproportionately absorbed into previously learned old classes, a boundary-level failure we call \textit{new-tail collapse}. We diagnose this failure with an old-new margin and analyze it through a boundary-risk view, where CLIP semantic overlap, sparse tail supervision, and old-class expansion jointly increase old-dominant violations. Motivated by this view, we propose Smooth Boundary-Risk Loss (SBRL), which estimates effective old rivals from CLIP{'}s vision-language geometry and suppresses their smooth old-side pressure against new-tail samples. MC-SMERG further consolidates task-wise adapters into a single efficient model. Across ImageNetSubset-LT, CIFAR100-LT, and Food101-LT, our method consistently outperforms strong CLIP-based CIL baselines and substantially reduces boundary violations in high-similarity, low-shot regimes."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lee-etal-2026-mitigating">
<titleInfo>
<title>Mitigating New-Tail Collapse in Long-Tailed Class-Incremental Learning with CLIP</title>
</titleInfo>
<name type="personal">
<namePart type="given">HyeokJun</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sungmin</namePart>
<namePart type="family">Cha</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">YoungJoon</namePart>
<namePart type="family">Yoo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Pre-trained vision–language models such as CLIP provide strong priors for class-incremental learning (CIL), yet they degrade sharply under long-tailed class streams. We show that this degradation is not merely a byproduct of forgetting: in frozen-CLIP long-tailed CIL, newly introduced tail classes are disproportionately absorbed into previously learned old classes, a boundary-level failure we call new-tail collapse. We diagnose this failure with an old-new margin and analyze it through a boundary-risk view, where CLIP semantic overlap, sparse tail supervision, and old-class expansion jointly increase old-dominant violations. Motivated by this view, we propose Smooth Boundary-Risk Loss (SBRL), which estimates effective old rivals from CLIP’s vision-language geometry and suppresses their smooth old-side pressure against new-tail samples. MC-SMERG further consolidates task-wise adapters into a single efficient model. Across ImageNetSubset-LT, CIFAR100-LT, and Food101-LT, our method consistently outperforms strong CLIP-based CIL baselines and substantially reduces boundary violations in high-similarity, low-shot regimes.</abstract>
<identifier type="citekey">lee-etal-2026-mitigating</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.1962</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1962/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>41837</start>
<end>41856</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Mitigating New-Tail Collapse in Long-Tailed Class-Incremental Learning with CLIP
%A Lee, HyeokJun
%A Cha, Sungmin
%A Yoo, YoungJoon
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F lee-etal-2026-mitigating
%X Pre-trained vision–language models such as CLIP provide strong priors for class-incremental learning (CIL), yet they degrade sharply under long-tailed class streams. We show that this degradation is not merely a byproduct of forgetting: in frozen-CLIP long-tailed CIL, newly introduced tail classes are disproportionately absorbed into previously learned old classes, a boundary-level failure we call new-tail collapse. We diagnose this failure with an old-new margin and analyze it through a boundary-risk view, where CLIP semantic overlap, sparse tail supervision, and old-class expansion jointly increase old-dominant violations. Motivated by this view, we propose Smooth Boundary-Risk Loss (SBRL), which estimates effective old rivals from CLIP’s vision-language geometry and suppresses their smooth old-side pressure against new-tail samples. MC-SMERG further consolidates task-wise adapters into a single efficient model. Across ImageNetSubset-LT, CIFAR100-LT, and Food101-LT, our method consistently outperforms strong CLIP-based CIL baselines and substantially reduces boundary violations in high-similarity, low-shot regimes.
%R 10.18653/2026.findings-emnlp.1962
%U https://aclanthology.org/2026.findings-emnlp.1962/
%U https://doi.org/10.18653/2026.findings-emnlp.1962
%P 41837-41856
Markdown (Informal)
[Mitigating New-Tail Collapse in Long-Tailed Class-Incremental Learning with CLIP](https://aclanthology.org/2026.findings-emnlp.1962/) (Lee et al., Findings 2026)
ACL