@inproceedings{radu-etal-2025-uncertainty,
title = "Uncertainty in Semantic Language Modeling with {PIXELS}",
author = "Radu, Stefania and
Zullich, Marco and
Valdenegro-Toro, Matias",
editor = "Noidea, Noidea",
booktitle = "Proceedings of the 2nd Workshop on Uncertainty-Aware NLP (UncertaiNLP 2025)",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.uncertainlp-main.11/",
pages = "103--119",
ISBN = "979-8-89176-349-4",
abstract = "Pixel-based language models aim to solve the vocabulary bottleneck problem in language modeling, but the challenge of uncertainty quantification remains open. The novelty of this work consists of analysing uncertainty and confidence in pixel-based language models across 18 languages and 7 scripts, all part of 3 semantically challenging tasks. This is achieved through several methods such as Monte Carlo Dropout, Transformer Attention, and Ensemble Learning. The results suggest that pixel-based models underestimate uncertainty when reconstructing patches. The uncertainty is also influenced by the script, with Latin languages displaying lower uncertainty. The findings on ensemble learning show better performance when applying hyperparameter tuning during the named entity recognition and question-answering tasks across 16 languages."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="radu-etal-2025-uncertainty">
<titleInfo>
<title>Uncertainty in Semantic Language Modeling with PIXELS</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stefania</namePart>
<namePart type="family">Radu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marco</namePart>
<namePart type="family">Zullich</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Matias</namePart>
<namePart type="family">Valdenegro-Toro</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2nd Workshop on Uncertainty-Aware NLP (UncertaiNLP 2025)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Noidea</namePart>
<namePart type="family">Noidea</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-349-4</identifier>
</relatedItem>
<abstract>Pixel-based language models aim to solve the vocabulary bottleneck problem in language modeling, but the challenge of uncertainty quantification remains open. The novelty of this work consists of analysing uncertainty and confidence in pixel-based language models across 18 languages and 7 scripts, all part of 3 semantically challenging tasks. This is achieved through several methods such as Monte Carlo Dropout, Transformer Attention, and Ensemble Learning. The results suggest that pixel-based models underestimate uncertainty when reconstructing patches. The uncertainty is also influenced by the script, with Latin languages displaying lower uncertainty. The findings on ensemble learning show better performance when applying hyperparameter tuning during the named entity recognition and question-answering tasks across 16 languages.</abstract>
<identifier type="citekey">radu-etal-2025-uncertainty</identifier>
<location>
<url>https://aclanthology.org/2025.uncertainlp-main.11/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>103</start>
<end>119</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Uncertainty in Semantic Language Modeling with PIXELS
%A Radu, Stefania
%A Zullich, Marco
%A Valdenegro-Toro, Matias
%Y Noidea, Noidea
%S Proceedings of the 2nd Workshop on Uncertainty-Aware NLP (UncertaiNLP 2025)
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-349-4
%F radu-etal-2025-uncertainty
%X Pixel-based language models aim to solve the vocabulary bottleneck problem in language modeling, but the challenge of uncertainty quantification remains open. The novelty of this work consists of analysing uncertainty and confidence in pixel-based language models across 18 languages and 7 scripts, all part of 3 semantically challenging tasks. This is achieved through several methods such as Monte Carlo Dropout, Transformer Attention, and Ensemble Learning. The results suggest that pixel-based models underestimate uncertainty when reconstructing patches. The uncertainty is also influenced by the script, with Latin languages displaying lower uncertainty. The findings on ensemble learning show better performance when applying hyperparameter tuning during the named entity recognition and question-answering tasks across 16 languages.
%U https://aclanthology.org/2025.uncertainlp-main.11/
%P 103-119
Markdown (Informal)
[Uncertainty in Semantic Language Modeling with PIXELS](https://aclanthology.org/2025.uncertainlp-main.11/) (Radu et al., UncertaiNLP 2025)
ACL
- Stefania Radu, Marco Zullich, and Matias Valdenegro-Toro. 2025. Uncertainty in Semantic Language Modeling with PIXELS. In Proceedings of the 2nd Workshop on Uncertainty-Aware NLP (UncertaiNLP 2025), pages 103–119, Suzhou, China. Association for Computational Linguistics.