@inproceedings{nishida-etal-2025-long,
title = "Long-Tail Crisis in Nearest Neighbor Language Models",
author = "Nishida, Yuto and
Morishita, Makoto and
Deguchi, Hiroyuki and
Kamigaito, Hidetaka and
Watanabe, Taro",
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.331/",
doi = "10.18653/v1/2025.findings-naacl.331",
pages = "5965--5978",
ISBN = "979-8-89176-195-7",
abstract = "The $k$-nearest-neighbor language model ($k$NN-LM), one of the retrieval-augmented language models, improves the perplexity for given text by directly accessing a large datastore built from any text data during inference.A widely held hypothesis for the success of $k$NN-LM is that its explicit memory, i.e., the datastore, enhances predictions for long-tail phenomena.However, prior works have primarily shown its ability to retrieve long-tail contexts, leaving the model{'}s performance remain underexplored in estimating the probabilities of long-tail target tokens during inference.In this paper, we investigate the behavior of $k$NN-LM on low-frequency tokens, examining prediction probability, retrieval accuracy, and token distribution in the datastore.Our experimental results reveal that $k$NN-LM does not improve prediction performance for low-frequency tokens but mainly benefits high-frequency tokens regardless of long-tail contexts in the datastore."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="nishida-etal-2025-long">
<titleInfo>
<title>Long-Tail Crisis in Nearest Neighbor Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuto</namePart>
<namePart type="family">Nishida</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Makoto</namePart>
<namePart type="family">Morishita</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hiroyuki</namePart>
<namePart type="family">Deguchi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hidetaka</namePart>
<namePart type="family">Kamigaito</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Taro</namePart>
<namePart type="family">Watanabe</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>The k-nearest-neighbor language model (kNN-LM), one of the retrieval-augmented language models, improves the perplexity for given text by directly accessing a large datastore built from any text data during inference.A widely held hypothesis for the success of kNN-LM is that its explicit memory, i.e., the datastore, enhances predictions for long-tail phenomena.However, prior works have primarily shown its ability to retrieve long-tail contexts, leaving the model’s performance remain underexplored in estimating the probabilities of long-tail target tokens during inference.In this paper, we investigate the behavior of kNN-LM on low-frequency tokens, examining prediction probability, retrieval accuracy, and token distribution in the datastore.Our experimental results reveal that kNN-LM does not improve prediction performance for low-frequency tokens but mainly benefits high-frequency tokens regardless of long-tail contexts in the datastore.</abstract>
<identifier type="citekey">nishida-etal-2025-long</identifier>
<identifier type="doi">10.18653/v1/2025.findings-naacl.331</identifier>
<location>
<url>https://aclanthology.org/2025.findings-naacl.331/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>5965</start>
<end>5978</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Long-Tail Crisis in Nearest Neighbor Language Models
%A Nishida, Yuto
%A Morishita, Makoto
%A Deguchi, Hiroyuki
%A Kamigaito, Hidetaka
%A Watanabe, Taro
%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 nishida-etal-2025-long
%X The k-nearest-neighbor language model (kNN-LM), one of the retrieval-augmented language models, improves the perplexity for given text by directly accessing a large datastore built from any text data during inference.A widely held hypothesis for the success of kNN-LM is that its explicit memory, i.e., the datastore, enhances predictions for long-tail phenomena.However, prior works have primarily shown its ability to retrieve long-tail contexts, leaving the model’s performance remain underexplored in estimating the probabilities of long-tail target tokens during inference.In this paper, we investigate the behavior of kNN-LM on low-frequency tokens, examining prediction probability, retrieval accuracy, and token distribution in the datastore.Our experimental results reveal that kNN-LM does not improve prediction performance for low-frequency tokens but mainly benefits high-frequency tokens regardless of long-tail contexts in the datastore.
%R 10.18653/v1/2025.findings-naacl.331
%U https://aclanthology.org/2025.findings-naacl.331/
%U https://doi.org/10.18653/v1/2025.findings-naacl.331
%P 5965-5978
Markdown (Informal)
[Long-Tail Crisis in Nearest Neighbor Language Models](https://aclanthology.org/2025.findings-naacl.331/) (Nishida et al., Findings 2025)
ACL
- Yuto Nishida, Makoto Morishita, Hiroyuki Deguchi, Hidetaka Kamigaito, and Taro Watanabe. 2025. Long-Tail Crisis in Nearest Neighbor Language Models. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 5965–5978, Albuquerque, New Mexico. Association for Computational Linguistics.