@inproceedings{abdullah-2026-im,
title = "{I}{'}m Sorry, but {I} Can{'}t Help with {B}raille: Revealing Accessibility Failures in State-of-the-Art {LLM}s",
author = "Abdullah",
editor = "Chakravarthi, Bharathi Raja and
B, Bharathi and
Buitelaar, Paul and
Thenmozhi, Durairaj and
Garc{\'i}a Cumbreras, Miguel {\'A}ngel and
Jim{\'e}nez Zafra, Salud Mar{\'i}a",
booktitle = "Proceedings of the Sixth Workshop on Language Technology for Equality, Diversity, Inclusion",
month = jul,
year = "2026",
address = "Virtual (Online)",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.ltedi-1.8/",
pages = "91--98",
ISBN = "979-8-89176-424-8",
abstract = "Large Language Models (LLMs) perform strongly on many language tasks, but their capability in structurally constrained, accessibility-critical modalities such as Braille remains unclear. We evaluate state-of-the-art LLMs on bidirectional Korean{--}Braille translation using a human-annotated dataset. Despite expectations that multilingual, instruction-tuned models can generalize to Braille via text representations, we find consistently poor, unstable outputs and substantial disagreement with human judgments. These results point to missing Braille-aware tokenization and weak alignment between Korean and Braille patterns. In contrast, supervised fine-tuning of a small model (T5-small) on the same data yields large and stable gains over zero-shot and prompted LLM baselines across standard metrics (SacreBLEU, ChrF++, CER, BLEU, ROUGE-L, METEOR, CIDEr). Our findings reveal a systematic limitation of current LLMs and demonstrate the effectiveness of modest task-specific supervision."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="abdullah-2026-im">
<titleInfo>
<title>I’m Sorry, but I Can’t Help with Braille: Revealing Accessibility Failures in State-of-the-Art LLMs</title>
</titleInfo>
<name>
<namePart>Abdullah</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 Sixth Workshop on Language Technology for Equality, Diversity, Inclusion</title>
</titleInfo>
<name type="personal">
<namePart type="given">Bharathi</namePart>
<namePart type="given">Raja</namePart>
<namePart type="family">Chakravarthi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bharathi</namePart>
<namePart type="family">B</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paul</namePart>
<namePart type="family">Buitelaar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Durairaj</namePart>
<namePart type="family">Thenmozhi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Miguel</namePart>
<namePart type="given">Ángel</namePart>
<namePart type="family">García Cumbreras</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Salud</namePart>
<namePart type="given">María</namePart>
<namePart type="family">Jiménez Zafra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Virtual (Online)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-424-8</identifier>
</relatedItem>
<abstract>Large Language Models (LLMs) perform strongly on many language tasks, but their capability in structurally constrained, accessibility-critical modalities such as Braille remains unclear. We evaluate state-of-the-art LLMs on bidirectional Korean–Braille translation using a human-annotated dataset. Despite expectations that multilingual, instruction-tuned models can generalize to Braille via text representations, we find consistently poor, unstable outputs and substantial disagreement with human judgments. These results point to missing Braille-aware tokenization and weak alignment between Korean and Braille patterns. In contrast, supervised fine-tuning of a small model (T5-small) on the same data yields large and stable gains over zero-shot and prompted LLM baselines across standard metrics (SacreBLEU, ChrF++, CER, BLEU, ROUGE-L, METEOR, CIDEr). Our findings reveal a systematic limitation of current LLMs and demonstrate the effectiveness of modest task-specific supervision.</abstract>
<identifier type="citekey">abdullah-2026-im</identifier>
<location>
<url>https://aclanthology.org/2026.ltedi-1.8/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>91</start>
<end>98</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T I’m Sorry, but I Can’t Help with Braille: Revealing Accessibility Failures in State-of-the-Art LLMs
%Y Chakravarthi, Bharathi Raja
%Y B, Bharathi
%Y Buitelaar, Paul
%Y Thenmozhi, Durairaj
%Y García Cumbreras, Miguel Ángel
%Y Jiménez Zafra, Salud María
%A Abdullah
%S Proceedings of the Sixth Workshop on Language Technology for Equality, Diversity, Inclusion
%D 2026
%8 July
%I Association for Computational Linguistics
%C Virtual (Online)
%@ 979-8-89176-424-8
%F abdullah-2026-im
%X Large Language Models (LLMs) perform strongly on many language tasks, but their capability in structurally constrained, accessibility-critical modalities such as Braille remains unclear. We evaluate state-of-the-art LLMs on bidirectional Korean–Braille translation using a human-annotated dataset. Despite expectations that multilingual, instruction-tuned models can generalize to Braille via text representations, we find consistently poor, unstable outputs and substantial disagreement with human judgments. These results point to missing Braille-aware tokenization and weak alignment between Korean and Braille patterns. In contrast, supervised fine-tuning of a small model (T5-small) on the same data yields large and stable gains over zero-shot and prompted LLM baselines across standard metrics (SacreBLEU, ChrF++, CER, BLEU, ROUGE-L, METEOR, CIDEr). Our findings reveal a systematic limitation of current LLMs and demonstrate the effectiveness of modest task-specific supervision.
%U https://aclanthology.org/2026.ltedi-1.8/
%P 91-98
Markdown (Informal)
[I’m Sorry, but I Can’t Help with Braille: Revealing Accessibility Failures in State-of-the-Art LLMs](https://aclanthology.org/2026.ltedi-1.8/) (Abdullah, LTEDI 2026)
ACL