@inproceedings{borchmann-etal-2025-arctic,
title = "Arctic-{TILT}. Business Document Understanding at Sub-Billion Scale",
author = "Borchmann, {\L}ukasz and
Pietruszka, Micha{\l} and
Ja{\'s}kowski, Wojciech and
Jurkiewicz, Dawid and
Halama, Piotr and
J{\'o}ziak, Pawe{\l} and
Garncarek, {\L}ukasz and
Liskowski, Pawe{\l} and
Szyndler, Karolina and
Gretkowski, Andrzej and
O{\l}tusek, Julita and
Nowakowska, Gabriela and
Zaw{\l}ocki, Artur and
Duhr, {\L}ukasz and
Dyda, Pawe{\l} and
Turski, Micha{\l}",
editor = "Rehm, Georg and
Li, Yunyao",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-industry.20/",
doi = "10.18653/v1/2025.acl-industry.20",
pages = "264--283",
ISBN = "979-8-89176-288-6",
abstract = "The vast portion of workloads employing LLMs involves answering questions grounded on PDF or scanned content. We introduce the Arctic-TILT achieving accuracy on par with models 1000$\times$ its size on these use cases. It can be finetuned and deployed on a single 24GB GPU, lowering operational costs while processing rich documents with up to 400k tokens. The model establishes state-of-the-art results on seven diverse Document Understanding benchmarks, as well as provides reliable confidence scores and quick inference, essential for processing files in large-scale or time-sensitive enterprise environments. We release Arctic-TILT weights and an efficient vLLM-based implementation on a permissive license."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="borchmann-etal-2025-arctic">
<titleInfo>
<title>Arctic-TILT. Business Document Understanding at Sub-Billion Scale</title>
</titleInfo>
<name type="personal">
<namePart type="given">Łukasz</namePart>
<namePart type="family">Borchmann</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michał</namePart>
<namePart type="family">Pietruszka</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wojciech</namePart>
<namePart type="family">Jaśkowski</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dawid</namePart>
<namePart type="family">Jurkiewicz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Piotr</namePart>
<namePart type="family">Halama</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paweł</namePart>
<namePart type="family">Józiak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Łukasz</namePart>
<namePart type="family">Garncarek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paweł</namePart>
<namePart type="family">Liskowski</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Karolina</namePart>
<namePart type="family">Szyndler</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrzej</namePart>
<namePart type="family">Gretkowski</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Julita</namePart>
<namePart type="family">Ołtusek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gabriela</namePart>
<namePart type="family">Nowakowska</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Artur</namePart>
<namePart type="family">Zawłocki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Łukasz</namePart>
<namePart type="family">Duhr</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paweł</namePart>
<namePart type="family">Dyda</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michał</namePart>
<namePart type="family">Turski</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Georg</namePart>
<namePart type="family">Rehm</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yunyao</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-288-6</identifier>
</relatedItem>
<abstract>The vast portion of workloads employing LLMs involves answering questions grounded on PDF or scanned content. We introduce the Arctic-TILT achieving accuracy on par with models 1000\times its size on these use cases. It can be finetuned and deployed on a single 24GB GPU, lowering operational costs while processing rich documents with up to 400k tokens. The model establishes state-of-the-art results on seven diverse Document Understanding benchmarks, as well as provides reliable confidence scores and quick inference, essential for processing files in large-scale or time-sensitive enterprise environments. We release Arctic-TILT weights and an efficient vLLM-based implementation on a permissive license.</abstract>
<identifier type="citekey">borchmann-etal-2025-arctic</identifier>
<identifier type="doi">10.18653/v1/2025.acl-industry.20</identifier>
<location>
<url>https://aclanthology.org/2025.acl-industry.20/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>264</start>
<end>283</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Arctic-TILT. Business Document Understanding at Sub-Billion Scale
%A Borchmann, Łukasz
%A Pietruszka, Michał
%A Jaśkowski, Wojciech
%A Jurkiewicz, Dawid
%A Halama, Piotr
%A Józiak, Paweł
%A Garncarek, Łukasz
%A Liskowski, Paweł
%A Szyndler, Karolina
%A Gretkowski, Andrzej
%A Ołtusek, Julita
%A Nowakowska, Gabriela
%A Zawłocki, Artur
%A Duhr, Łukasz
%A Dyda, Paweł
%A Turski, Michał
%Y Rehm, Georg
%Y Li, Yunyao
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-288-6
%F borchmann-etal-2025-arctic
%X The vast portion of workloads employing LLMs involves answering questions grounded on PDF or scanned content. We introduce the Arctic-TILT achieving accuracy on par with models 1000\times its size on these use cases. It can be finetuned and deployed on a single 24GB GPU, lowering operational costs while processing rich documents with up to 400k tokens. The model establishes state-of-the-art results on seven diverse Document Understanding benchmarks, as well as provides reliable confidence scores and quick inference, essential for processing files in large-scale or time-sensitive enterprise environments. We release Arctic-TILT weights and an efficient vLLM-based implementation on a permissive license.
%R 10.18653/v1/2025.acl-industry.20
%U https://aclanthology.org/2025.acl-industry.20/
%U https://doi.org/10.18653/v1/2025.acl-industry.20
%P 264-283
Markdown (Informal)
[Arctic-TILT. Business Document Understanding at Sub-Billion Scale](https://aclanthology.org/2025.acl-industry.20/) (Borchmann et al., ACL 2025)
ACL
- Łukasz Borchmann, Michał Pietruszka, Wojciech Jaśkowski, Dawid Jurkiewicz, Piotr Halama, Paweł Józiak, Łukasz Garncarek, Paweł Liskowski, Karolina Szyndler, Andrzej Gretkowski, Julita Ołtusek, Gabriela Nowakowska, Artur Zawłocki, Łukasz Duhr, Paweł Dyda, and Michał Turski. 2025. Arctic-TILT. Business Document Understanding at Sub-Billion Scale. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track), pages 264–283, Vienna, Austria. Association for Computational Linguistics.