@inproceedings{devalla-etal-2026-indicdetect,
title = "{I}ndic{D}etect: Evaluating Cross-Lingual {LLM}-Generated Text Detection for {H}indi, {T}elugu, and {T}amil",
author = "Devalla, Bhaskar Ganesh and
Wu, Junchao and
Dokuparthi, Nilesh and
Yaluru, Greeshma and
Rodriguez, Tatiana Muniz and
Chao, Lidia S. and
Wong, Derek F.",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.885/",
pages = "18683--18731",
ISBN = "979-8-89176-462-0",
abstract = "The rapid proliferation of LLMs has further heightened the necessity to develop dependable AI-generated text detection, especially beyond English. Nevertheless, current benchmarks pay little attention to Indic languages and test detectors in idealized settings that do not represent the real world. We present a generalized benchmark for AI-generated text detection in Hindi, Telugu, and Tamil, which we name IndicDetect is a benchmark designed to assess the robustness of detectors under realistic distribution shifts. IndicDetect comprises highly curated human-written texts matched with LLM-generated counterparts across various domains and generators, and systematically evaluates detectors in the presence of domain shift, generator shift, and adversarial perturbation. Using a single and repeatable evaluation scheme, we evaluate a wide range of statistical and neural detectors. We find substantial robustness failures: supervised neural detectors perform well in-distribution, but most training-free methods fail catastrophically on unseen generators and adversarial attacks, with larger failures observed for Telugu and Tamil than for Hindi. These results highlight that the primary weakness of existing detectors in Indic settings lies in their robustness, not peak accuracy. IndicDetect provides standard data splits, an evaluation protocol, and baselines to establish a robust, language-aware foundation for AI-generated text detection in Indic scripts."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="devalla-etal-2026-indicdetect">
<titleInfo>
<title>IndicDetect: Evaluating Cross-Lingual LLM-Generated Text Detection for Hindi, Telugu, and Tamil</title>
</titleInfo>
<name type="personal">
<namePart type="given">Bhaskar</namePart>
<namePart type="given">Ganesh</namePart>
<namePart type="family">Devalla</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junchao</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nilesh</namePart>
<namePart type="family">Dokuparthi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Greeshma</namePart>
<namePart type="family">Yaluru</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tatiana</namePart>
<namePart type="given">Muniz</namePart>
<namePart type="family">Rodriguez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lidia</namePart>
<namePart type="given">S</namePart>
<namePart type="family">Chao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Derek</namePart>
<namePart type="given">F</namePart>
<namePart type="family">Wong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>The rapid proliferation of LLMs has further heightened the necessity to develop dependable AI-generated text detection, especially beyond English. Nevertheless, current benchmarks pay little attention to Indic languages and test detectors in idealized settings that do not represent the real world. We present a generalized benchmark for AI-generated text detection in Hindi, Telugu, and Tamil, which we name IndicDetect is a benchmark designed to assess the robustness of detectors under realistic distribution shifts. IndicDetect comprises highly curated human-written texts matched with LLM-generated counterparts across various domains and generators, and systematically evaluates detectors in the presence of domain shift, generator shift, and adversarial perturbation. Using a single and repeatable evaluation scheme, we evaluate a wide range of statistical and neural detectors. We find substantial robustness failures: supervised neural detectors perform well in-distribution, but most training-free methods fail catastrophically on unseen generators and adversarial attacks, with larger failures observed for Telugu and Tamil than for Hindi. These results highlight that the primary weakness of existing detectors in Indic settings lies in their robustness, not peak accuracy. IndicDetect provides standard data splits, an evaluation protocol, and baselines to establish a robust, language-aware foundation for AI-generated text detection in Indic scripts.</abstract>
<identifier type="citekey">devalla-etal-2026-indicdetect</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.885/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>18683</start>
<end>18731</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T IndicDetect: Evaluating Cross-Lingual LLM-Generated Text Detection for Hindi, Telugu, and Tamil
%A Devalla, Bhaskar Ganesh
%A Wu, Junchao
%A Dokuparthi, Nilesh
%A Yaluru, Greeshma
%A Rodriguez, Tatiana Muniz
%A Chao, Lidia S.
%A Wong, Derek F.
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F devalla-etal-2026-indicdetect
%X The rapid proliferation of LLMs has further heightened the necessity to develop dependable AI-generated text detection, especially beyond English. Nevertheless, current benchmarks pay little attention to Indic languages and test detectors in idealized settings that do not represent the real world. We present a generalized benchmark for AI-generated text detection in Hindi, Telugu, and Tamil, which we name IndicDetect is a benchmark designed to assess the robustness of detectors under realistic distribution shifts. IndicDetect comprises highly curated human-written texts matched with LLM-generated counterparts across various domains and generators, and systematically evaluates detectors in the presence of domain shift, generator shift, and adversarial perturbation. Using a single and repeatable evaluation scheme, we evaluate a wide range of statistical and neural detectors. We find substantial robustness failures: supervised neural detectors perform well in-distribution, but most training-free methods fail catastrophically on unseen generators and adversarial attacks, with larger failures observed for Telugu and Tamil than for Hindi. These results highlight that the primary weakness of existing detectors in Indic settings lies in their robustness, not peak accuracy. IndicDetect provides standard data splits, an evaluation protocol, and baselines to establish a robust, language-aware foundation for AI-generated text detection in Indic scripts.
%U https://aclanthology.org/2026.findings-emnlp.885/
%P 18683-18731
Markdown (Informal)
[IndicDetect: Evaluating Cross-Lingual LLM-Generated Text Detection for Hindi, Telugu, and Tamil](https://aclanthology.org/2026.findings-emnlp.885/) (Devalla et al., Findings 2026)
ACL
- Bhaskar Ganesh Devalla, Junchao Wu, Nilesh Dokuparthi, Greeshma Yaluru, Tatiana Muniz Rodriguez, Lidia S. Chao, and Derek F. Wong. 2026. IndicDetect: Evaluating Cross-Lingual LLM-Generated Text Detection for Hindi, Telugu, and Tamil. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 18683–18731, Budapest, Hungary. Association for Computational Linguistics.