@inproceedings{cao-etal-2026-optimal,
title = "Optimal Transport Depth Up-Scaling",
author = "Cao, Mingzi and
Wang, Xi and
Aletras, Nikolaos",
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.520/",
pages = "10898--10913",
ISBN = "979-8-89176-462-0",
abstract = "Pre-training Large Language Models (LLMs) from scratch at larger scales yields remarkable performance but incurs substantially high training costs. Depth up-scaling provides an efficient alternative by inserting new layers into a pre-trained LLM, avoiding training from scratch. However, most existing methods copying or averaging base layers for new layer, which misalign functionally corresponding neurons, leading to neuron permutation mismatch that harms performance. To address this issue, we propose Optimal Transport Depth Up-Scaling (OT-DUS), which leverages Optimal Transport (OT) theory to align and fuse functionally corresponding neurons module by module in adjacent base layers for new layer construction. OT-DUS achieves better overall performance in both general and specialized domains than existing methods for continual pre-training and supervised fine-tuning across different model sizes and model families. Our analysis of insertion strategies further finds that inserting new layers at higher positions yields not only stronger performance but also improved training efficiency."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="cao-etal-2026-optimal">
<titleInfo>
<title>Optimal Transport Depth Up-Scaling</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mingzi</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xi</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nikolaos</namePart>
<namePart type="family">Aletras</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>Pre-training Large Language Models (LLMs) from scratch at larger scales yields remarkable performance but incurs substantially high training costs. Depth up-scaling provides an efficient alternative by inserting new layers into a pre-trained LLM, avoiding training from scratch. However, most existing methods copying or averaging base layers for new layer, which misalign functionally corresponding neurons, leading to neuron permutation mismatch that harms performance. To address this issue, we propose Optimal Transport Depth Up-Scaling (OT-DUS), which leverages Optimal Transport (OT) theory to align and fuse functionally corresponding neurons module by module in adjacent base layers for new layer construction. OT-DUS achieves better overall performance in both general and specialized domains than existing methods for continual pre-training and supervised fine-tuning across different model sizes and model families. Our analysis of insertion strategies further finds that inserting new layers at higher positions yields not only stronger performance but also improved training efficiency.</abstract>
<identifier type="citekey">cao-etal-2026-optimal</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.520/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>10898</start>
<end>10913</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Optimal Transport Depth Up-Scaling
%A Cao, Mingzi
%A Wang, Xi
%A Aletras, Nikolaos
%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 cao-etal-2026-optimal
%X Pre-training Large Language Models (LLMs) from scratch at larger scales yields remarkable performance but incurs substantially high training costs. Depth up-scaling provides an efficient alternative by inserting new layers into a pre-trained LLM, avoiding training from scratch. However, most existing methods copying or averaging base layers for new layer, which misalign functionally corresponding neurons, leading to neuron permutation mismatch that harms performance. To address this issue, we propose Optimal Transport Depth Up-Scaling (OT-DUS), which leverages Optimal Transport (OT) theory to align and fuse functionally corresponding neurons module by module in adjacent base layers for new layer construction. OT-DUS achieves better overall performance in both general and specialized domains than existing methods for continual pre-training and supervised fine-tuning across different model sizes and model families. Our analysis of insertion strategies further finds that inserting new layers at higher positions yields not only stronger performance but also improved training efficiency.
%U https://aclanthology.org/2026.findings-emnlp.520/
%P 10898-10913
Markdown (Informal)
[Optimal Transport Depth Up-Scaling](https://aclanthology.org/2026.findings-emnlp.520/) (Cao et al., Findings 2026)
ACL
- Mingzi Cao, Xi Wang, and Nikolaos Aletras. 2026. Optimal Transport Depth Up-Scaling. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 10898–10913, Budapest, Hungary. Association for Computational Linguistics.