@inproceedings{su-etal-2026-improves,
title = "What Improves During Pretraining? Tracking Knowledge Acquisition with Point-in-Time Models",
author = "Su, Chang and
Wu, Bin and
Yilmaz, Emine and
Lu, Yao",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the {A}sia-Pacific Chapter of the {A}ssociation for {C}omputational {L}inguistics (Volume 2: Short Papers)",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.aacl-short.56/",
pages = "879--891",
ISBN = "979-8-89176-440-8",
abstract = "Language models pretrained on web corpora collected over multiple years learn both general and domain-specific knowledge. How these different kinds of knowledge are learned as pretraining progresses remains unclear. Recent point-in-time models trained on web snapshots from different years provide a way to probe how knowledge prevalence varies over time. Using these models, we propose a temporal variability score (TVS), defined as the coefficient of variation of a fact{'}s negative log-likelihood across models. A higher score suggests that the knowledge is less consistently represented across years of web text. We evaluate six benchmarks covering diverse topics using six models from multiple open-source model families. We find that low- and high-TVS items are learned similarly early in training, but high-TVS items improve faster later on. At the end of the pretraining run, high-TVS items outperform low-TVS items in 34 of 36 model{--}benchmark settings, with an average advantage of 7.3 points of chance-normalized accuracy. This pattern is especially pronounced on knowledge-intensive tasks, even when comparing items of similar difficulty. Our findings reveal differences in how knowledge is learned over the course of pretraining, with temporal variability providing a diagnostic of these learning dynamics beyond aggregate benchmark scores."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="su-etal-2026-improves">
<titleInfo>
<title>What Improves During Pretraining? Tracking Knowledge Acquisition with Point-in-Time Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Chang</namePart>
<namePart type="family">Su</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bin</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Emine</namePart>
<namePart type="family">Yilmaz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yao</namePart>
<namePart type="family">Lu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-440-8</identifier>
</relatedItem>
<abstract>Language models pretrained on web corpora collected over multiple years learn both general and domain-specific knowledge. How these different kinds of knowledge are learned as pretraining progresses remains unclear. Recent point-in-time models trained on web snapshots from different years provide a way to probe how knowledge prevalence varies over time. Using these models, we propose a temporal variability score (TVS), defined as the coefficient of variation of a fact’s negative log-likelihood across models. A higher score suggests that the knowledge is less consistently represented across years of web text. We evaluate six benchmarks covering diverse topics using six models from multiple open-source model families. We find that low- and high-TVS items are learned similarly early in training, but high-TVS items improve faster later on. At the end of the pretraining run, high-TVS items outperform low-TVS items in 34 of 36 model–benchmark settings, with an average advantage of 7.3 points of chance-normalized accuracy. This pattern is especially pronounced on knowledge-intensive tasks, even when comparing items of similar difficulty. Our findings reveal differences in how knowledge is learned over the course of pretraining, with temporal variability providing a diagnostic of these learning dynamics beyond aggregate benchmark scores.</abstract>
<identifier type="citekey">su-etal-2026-improves</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-short.56/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>879</start>
<end>891</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T What Improves During Pretraining? Tracking Knowledge Acquisition with Point-in-Time Models
%A Su, Chang
%A Wu, Bin
%A Yilmaz, Emine
%A Lu, Yao
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-440-8
%F su-etal-2026-improves
%X Language models pretrained on web corpora collected over multiple years learn both general and domain-specific knowledge. How these different kinds of knowledge are learned as pretraining progresses remains unclear. Recent point-in-time models trained on web snapshots from different years provide a way to probe how knowledge prevalence varies over time. Using these models, we propose a temporal variability score (TVS), defined as the coefficient of variation of a fact’s negative log-likelihood across models. A higher score suggests that the knowledge is less consistently represented across years of web text. We evaluate six benchmarks covering diverse topics using six models from multiple open-source model families. We find that low- and high-TVS items are learned similarly early in training, but high-TVS items improve faster later on. At the end of the pretraining run, high-TVS items outperform low-TVS items in 34 of 36 model–benchmark settings, with an average advantage of 7.3 points of chance-normalized accuracy. This pattern is especially pronounced on knowledge-intensive tasks, even when comparing items of similar difficulty. Our findings reveal differences in how knowledge is learned over the course of pretraining, with temporal variability providing a diagnostic of these learning dynamics beyond aggregate benchmark scores.
%U https://aclanthology.org/2026.aacl-short.56/
%P 879-891
Markdown (Informal)
[What Improves During Pretraining? Tracking Knowledge Acquisition with Point-in-Time Models](https://aclanthology.org/2026.aacl-short.56/) (Su et al., AACL-IJCNLP 2026)
ACL