@inproceedings{stachura-2025-perplexity,
title = "Perplexity-Driven Contrastive Scoring for Unsupervised Detection of {AI}-Generated Texts in {P}olish",
author = "Stachura, Damian",
editor = "Kobyli{\'n}ski, {\L}ukasz and
Wr{\'o}blewska, Alina and
Ogrodniczuk, Maciej",
booktitle = "Proceedings of the {P}ol{E}val 2025 Workshop",
month = nov,
year = "2025",
address = "Warsaw",
publisher = "Institute of Computer Science PAS and Association for Computational Linguistics",
url = "https://aclanthology.org/2025.poleval-main.4/",
pages = "21--25",
abstract = "The SMIGIEL competition at PolEval 2025 focuses on distinguishing Polish human-written text from AI-generated text. I participated in one of the subtasks that required a zero-shot detection method. My solution adapts the Binoculars detector by pairing language models and using calibrated thresholds. Specifically, I replaced the English language models from the original Binoculars method with models trained on Polish corpora. This approach achieved first place in the chosen competition track. Overall, my findings demonstrate that domain-specific language models and careful thresholding enable state-of-the-art zero-shot AI-text detection performance across new languages and domains. The code is publicly available at https://github.com/damian1996/2025-smigiel."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="stachura-2025-perplexity">
<titleInfo>
<title>Perplexity-Driven Contrastive Scoring for Unsupervised Detection of AI-Generated Texts in Polish</title>
</titleInfo>
<name type="personal">
<namePart type="given">Damian</namePart>
<namePart type="family">Stachura</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 PolEval 2025 Workshop</title>
</titleInfo>
<name type="personal">
<namePart type="given">Łukasz</namePart>
<namePart type="family">Kobyliński</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alina</namePart>
<namePart type="family">Wróblewska</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maciej</namePart>
<namePart type="family">Ogrodniczuk</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Institute of Computer Science PAS and Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Warsaw</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>The SMIGIEL competition at PolEval 2025 focuses on distinguishing Polish human-written text from AI-generated text. I participated in one of the subtasks that required a zero-shot detection method. My solution adapts the Binoculars detector by pairing language models and using calibrated thresholds. Specifically, I replaced the English language models from the original Binoculars method with models trained on Polish corpora. This approach achieved first place in the chosen competition track. Overall, my findings demonstrate that domain-specific language models and careful thresholding enable state-of-the-art zero-shot AI-text detection performance across new languages and domains. The code is publicly available at https://github.com/damian1996/2025-smigiel.</abstract>
<identifier type="citekey">stachura-2025-perplexity</identifier>
<location>
<url>https://aclanthology.org/2025.poleval-main.4/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>21</start>
<end>25</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Perplexity-Driven Contrastive Scoring for Unsupervised Detection of AI-Generated Texts in Polish
%A Stachura, Damian
%Y Kobyliński, Łukasz
%Y Wróblewska, Alina
%Y Ogrodniczuk, Maciej
%S Proceedings of the PolEval 2025 Workshop
%D 2025
%8 November
%I Institute of Computer Science PAS and Association for Computational Linguistics
%C Warsaw
%F stachura-2025-perplexity
%X The SMIGIEL competition at PolEval 2025 focuses on distinguishing Polish human-written text from AI-generated text. I participated in one of the subtasks that required a zero-shot detection method. My solution adapts the Binoculars detector by pairing language models and using calibrated thresholds. Specifically, I replaced the English language models from the original Binoculars method with models trained on Polish corpora. This approach achieved first place in the chosen competition track. Overall, my findings demonstrate that domain-specific language models and careful thresholding enable state-of-the-art zero-shot AI-text detection performance across new languages and domains. The code is publicly available at https://github.com/damian1996/2025-smigiel.
%U https://aclanthology.org/2025.poleval-main.4/
%P 21-25
Markdown (Informal)
[Perplexity-Driven Contrastive Scoring for Unsupervised Detection of AI-Generated Texts in Polish](https://aclanthology.org/2025.poleval-main.4/) (Stachura, PolEval 2025)
ACL