@inproceedings{jacobs-grobol-2026-scaling,
title = "On the scaling relationship between cloze probabilities and language model next-token prediction",
author = "Jacobs, Cassandra L and
Grobol, Morgan",
editor = "Bonial, Claire and
Berzak, Yevgeni",
booktitle = "Proceedings of the 30th Conference on Computational Natural Language Learning",
month = jul,
year = "2026",
address = "San Diego, California, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.conll-main.32/",
pages = "544--554",
ISBN = "979-8-89176-410-1",
abstract = "Recent work has shown that larger language models have better predictive power for eye movement and reading time data. However, we know less about how model capacity relates to human production statistics in the cloze task, which are used to predict reading times as well. While even the best models under-allocate probability mass to human responses, larger models assign higher-quality estimates of next tokens and their likelihood of production in cloze data because they are less sensitive to lexical co-occurrence statistics while being better aligned semantically to human cloze responses. The results provide support for the claim that the greater memorization capacity of larger models helps them guess more semantically appropriate words, but makes them less sensitive to low-level information that is relevant for word recognition."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jacobs-grobol-2026-scaling">
<titleInfo>
<title>On the scaling relationship between cloze probabilities and language model next-token prediction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Cassandra</namePart>
<namePart type="given">L</namePart>
<namePart type="family">Jacobs</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Morgan</namePart>
<namePart type="family">Grobol</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 30th Conference on Computational Natural Language Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Claire</namePart>
<namePart type="family">Bonial</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yevgeni</namePart>
<namePart type="family">Berzak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-410-1</identifier>
</relatedItem>
<abstract>Recent work has shown that larger language models have better predictive power for eye movement and reading time data. However, we know less about how model capacity relates to human production statistics in the cloze task, which are used to predict reading times as well. While even the best models under-allocate probability mass to human responses, larger models assign higher-quality estimates of next tokens and their likelihood of production in cloze data because they are less sensitive to lexical co-occurrence statistics while being better aligned semantically to human cloze responses. The results provide support for the claim that the greater memorization capacity of larger models helps them guess more semantically appropriate words, but makes them less sensitive to low-level information that is relevant for word recognition.</abstract>
<identifier type="citekey">jacobs-grobol-2026-scaling</identifier>
<location>
<url>https://aclanthology.org/2026.conll-main.32/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>544</start>
<end>554</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T On the scaling relationship between cloze probabilities and language model next-token prediction
%A Jacobs, Cassandra L.
%A Grobol, Morgan
%Y Bonial, Claire
%Y Berzak, Yevgeni
%S Proceedings of the 30th Conference on Computational Natural Language Learning
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, USA
%@ 979-8-89176-410-1
%F jacobs-grobol-2026-scaling
%X Recent work has shown that larger language models have better predictive power for eye movement and reading time data. However, we know less about how model capacity relates to human production statistics in the cloze task, which are used to predict reading times as well. While even the best models under-allocate probability mass to human responses, larger models assign higher-quality estimates of next tokens and their likelihood of production in cloze data because they are less sensitive to lexical co-occurrence statistics while being better aligned semantically to human cloze responses. The results provide support for the claim that the greater memorization capacity of larger models helps them guess more semantically appropriate words, but makes them less sensitive to low-level information that is relevant for word recognition.
%U https://aclanthology.org/2026.conll-main.32/
%P 544-554
Markdown (Informal)
[On the scaling relationship between cloze probabilities and language model next-token prediction](https://aclanthology.org/2026.conll-main.32/) (Jacobs & Grobol, CoNLL 2026)
ACL