@inproceedings{korner-etal-2026-copy,
title = "Copy First, Translate Later: Interpreting Translation Dynamics in Multilingual Pretraining",
author = {K{\"o}rner, Felicia and
Matveev, Maria and
Eichin, Florian and
Kutyniok, Gitta and
Plank, Barbara and
Hedderich, Michael A.},
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2684/",
pages = "61018--61051",
ISBN = "979-8-89176-459-0",
abstract = "Large language models exhibit impressive cross-lingual capabilities. However, *how* cross-lingual generalization emerges has only been analyzed through isolated factors and at sparse points during training, limiting our understanding {---} particularly of the early phases of learning. To study the early trajectory of linguistic and translation capabilities, we pretrain a multilingual 1.7B model on nine diverse languages, capturing checkpoints at a much finer granularity. We use word-level translation as a testbed, introducing a novel dataset to trace how translation develops over training through behavioral analyses, model-component analysis, and parameter-based ablations. We find that the model quickly acquires basic linguistic capabilities in parallel with token-level copying, while translation develops in two distinct phases: an initial phase dominated by copying and surface-level similarities, and a second phase in which generalizing translation mechanisms develop and copying is refined. Together, these findings provide a fine-grained view of how cross-lingual generalization develops during multilingual pretraining."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="korner-etal-2026-copy">
<titleInfo>
<title>Copy First, Translate Later: Interpreting Translation Dynamics in Multilingual Pretraining</title>
</titleInfo>
<name type="personal">
<namePart type="given">Felicia</namePart>
<namePart type="family">Körner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Matveev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Florian</namePart>
<namePart type="family">Eichin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gitta</namePart>
<namePart type="family">Kutyniok</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Barbara</namePart>
<namePart type="family">Plank</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michael</namePart>
<namePart type="given">A</namePart>
<namePart type="family">Hedderich</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Large language models exhibit impressive cross-lingual capabilities. However, *how* cross-lingual generalization emerges has only been analyzed through isolated factors and at sparse points during training, limiting our understanding — particularly of the early phases of learning. To study the early trajectory of linguistic and translation capabilities, we pretrain a multilingual 1.7B model on nine diverse languages, capturing checkpoints at a much finer granularity. We use word-level translation as a testbed, introducing a novel dataset to trace how translation develops over training through behavioral analyses, model-component analysis, and parameter-based ablations. We find that the model quickly acquires basic linguistic capabilities in parallel with token-level copying, while translation develops in two distinct phases: an initial phase dominated by copying and surface-level similarities, and a second phase in which generalizing translation mechanisms develop and copying is refined. Together, these findings provide a fine-grained view of how cross-lingual generalization develops during multilingual pretraining.</abstract>
<identifier type="citekey">korner-etal-2026-copy</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2684/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>61018</start>
<end>61051</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Copy First, Translate Later: Interpreting Translation Dynamics in Multilingual Pretraining
%A Körner, Felicia
%A Matveev, Maria
%A Eichin, Florian
%A Kutyniok, Gitta
%A Plank, Barbara
%A Hedderich, Michael A.
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F korner-etal-2026-copy
%X Large language models exhibit impressive cross-lingual capabilities. However, *how* cross-lingual generalization emerges has only been analyzed through isolated factors and at sparse points during training, limiting our understanding — particularly of the early phases of learning. To study the early trajectory of linguistic and translation capabilities, we pretrain a multilingual 1.7B model on nine diverse languages, capturing checkpoints at a much finer granularity. We use word-level translation as a testbed, introducing a novel dataset to trace how translation develops over training through behavioral analyses, model-component analysis, and parameter-based ablations. We find that the model quickly acquires basic linguistic capabilities in parallel with token-level copying, while translation develops in two distinct phases: an initial phase dominated by copying and surface-level similarities, and a second phase in which generalizing translation mechanisms develop and copying is refined. Together, these findings provide a fine-grained view of how cross-lingual generalization develops during multilingual pretraining.
%U https://aclanthology.org/2026.emnlp-main.2684/
%P 61018-61051
Markdown (Informal)
[Copy First, Translate Later: Interpreting Translation Dynamics in Multilingual Pretraining](https://aclanthology.org/2026.emnlp-main.2684/) (Körner et al., EMNLP 2026)
ACL