@inproceedings{aizawa-etal-2026-building,
title = "Building Effective {J}apanese Medical {LLM}s with an Open Recipe for Domain Adaptation through Continued Pre-training",
author = "Aizawa, Akiko and
Arase, Yuki and
Cheng, Fei and
Huang, Jiahao and
Huang, Zhiyi and
Jiang, Junfeng and
Kanazawa, Teruhito and
Kawahara, Daisuke and
Kobayashi, Kazuma and
Kodama, Takashi and
Kurohashi, Sadao and
Oda, Yusuke and
Tsuta, Yuma and
Wan, Zhen and
Yang, Zhishen and
Yokota, Rio",
editor = "Piperidis, Stelios and
Bel, N{\'u}ria and
van den Heuvel, Henk and
Ide, Nancy and
Krek, Simon and
Toral, Antonio",
booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "ELRA Language Resource Association",
url = "https://aclanthology.org/2026.lrec-1.817/",
doi = "10.63317/47uvbxqph5ph",
pages = "10405--10423",
abstract = "In high-stakes domains such as medicine, ensuring transparency of the training corpus is essential, with careful consideration of local healthcare landscapes; however, the majority of existing medical large language models (LLMs) have not disclosed the details of their training corpora. Here, we introduce an open recipe for domain adaptation of LLMs to the Japanese medical domain. We employed fully open-source Japanese general-domain LLMs as base models, whose pre-training datasets are also disclosed. To establish effective corpora for domain adaptation through continued pre-training, we started with small-scale medical datasets and ultimately constructed a medical corpus consisting of 79.6B tokens, incorporating local clinical guidelines, medical textbooks, and other domain-specific resources. The resulting LLM from continued pre-training, namely SIP-med-llm-8x13B, with an active parameter count of 22B, demonstrated favorable accuracy on benchmarks including the Japanese National Medical Examination. This performance was comparable to that of 70B-parameter open-weight models whose construction details remain non-transparent. This represents the first case in the Japanese medical field where complete corpus details have been disclosed for fully from-scratch development, providing important insights for future efforts to construct medical LLMs tailored to the specific characteristics of local contexts. The model is available publicly at this Hugging Face repository: \url{https://huggingface.co/SIP-med-LLM/SIP-jmed-llm-2-8x13b-OP-instruct}."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="aizawa-etal-2026-building">
<titleInfo>
<title>Building Effective Japanese Medical LLMs with an Open Recipe for Domain Adaptation through Continued Pre-training</title>
</titleInfo>
<name type="personal">
<namePart type="given">Akiko</namePart>
<namePart type="family">Aizawa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuki</namePart>
<namePart type="family">Arase</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fei</namePart>
<namePart type="family">Cheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiahao</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhiyi</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junfeng</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Teruhito</namePart>
<namePart type="family">Kanazawa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Daisuke</namePart>
<namePart type="family">Kawahara</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kazuma</namePart>
<namePart type="family">Kobayashi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Takashi</namePart>
<namePart type="family">Kodama</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sadao</namePart>
<namePart type="family">Kurohashi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yusuke</namePart>
<namePart type="family">Oda</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuma</namePart>
<namePart type="family">Tsuta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhen</namePart>
<namePart type="family">Wan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhishen</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rio</namePart>
<namePart type="family">Yokota</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Fifteenth Language Resources and Evaluation Conference</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stelios</namePart>
<namePart type="family">Piperidis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Núria</namePart>
<namePart type="family">Bel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Henk</namePart>
<namePart type="family">van den Heuvel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nancy</namePart>
<namePart type="family">Ide</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Krek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Toral</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resource Association</publisher>
<place>
<placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>In high-stakes domains such as medicine, ensuring transparency of the training corpus is essential, with careful consideration of local healthcare landscapes; however, the majority of existing medical large language models (LLMs) have not disclosed the details of their training corpora. Here, we introduce an open recipe for domain adaptation of LLMs to the Japanese medical domain. We employed fully open-source Japanese general-domain LLMs as base models, whose pre-training datasets are also disclosed. To establish effective corpora for domain adaptation through continued pre-training, we started with small-scale medical datasets and ultimately constructed a medical corpus consisting of 79.6B tokens, incorporating local clinical guidelines, medical textbooks, and other domain-specific resources. The resulting LLM from continued pre-training, namely SIP-med-llm-8x13B, with an active parameter count of 22B, demonstrated favorable accuracy on benchmarks including the Japanese National Medical Examination. This performance was comparable to that of 70B-parameter open-weight models whose construction details remain non-transparent. This represents the first case in the Japanese medical field where complete corpus details have been disclosed for fully from-scratch development, providing important insights for future efforts to construct medical LLMs tailored to the specific characteristics of local contexts. The model is available publicly at this Hugging Face repository: https://huggingface.co/SIP-med-LLM/SIP-jmed-llm-2-8x13b-OP-instruct.</abstract>
<identifier type="citekey">aizawa-etal-2026-building</identifier>
<identifier type="doi">10.63317/47uvbxqph5ph</identifier>
<location>
<url>https://aclanthology.org/2026.lrec-1.817/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>10405</start>
<end>10423</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Building Effective Japanese Medical LLMs with an Open Recipe for Domain Adaptation through Continued Pre-training
%A Aizawa, Akiko
%A Arase, Yuki
%A Cheng, Fei
%A Huang, Jiahao
%A Huang, Zhiyi
%A Jiang, Junfeng
%A Kanazawa, Teruhito
%A Kawahara, Daisuke
%A Kobayashi, Kazuma
%A Kodama, Takashi
%A Kurohashi, Sadao
%A Oda, Yusuke
%A Tsuta, Yuma
%A Wan, Zhen
%A Yang, Zhishen
%A Yokota, Rio
%Y Piperidis, Stelios
%Y Bel, Núria
%Y van den Heuvel, Henk
%Y Ide, Nancy
%Y Krek, Simon
%Y Toral, Antonio
%S Proceedings of the Fifteenth Language Resources and Evaluation Conference
%D 2026
%8 May
%I ELRA Language Resource Association
%C Palma de Mallorca, Spain
%F aizawa-etal-2026-building
%X In high-stakes domains such as medicine, ensuring transparency of the training corpus is essential, with careful consideration of local healthcare landscapes; however, the majority of existing medical large language models (LLMs) have not disclosed the details of their training corpora. Here, we introduce an open recipe for domain adaptation of LLMs to the Japanese medical domain. We employed fully open-source Japanese general-domain LLMs as base models, whose pre-training datasets are also disclosed. To establish effective corpora for domain adaptation through continued pre-training, we started with small-scale medical datasets and ultimately constructed a medical corpus consisting of 79.6B tokens, incorporating local clinical guidelines, medical textbooks, and other domain-specific resources. The resulting LLM from continued pre-training, namely SIP-med-llm-8x13B, with an active parameter count of 22B, demonstrated favorable accuracy on benchmarks including the Japanese National Medical Examination. This performance was comparable to that of 70B-parameter open-weight models whose construction details remain non-transparent. This represents the first case in the Japanese medical field where complete corpus details have been disclosed for fully from-scratch development, providing important insights for future efforts to construct medical LLMs tailored to the specific characteristics of local contexts. The model is available publicly at this Hugging Face repository: https://huggingface.co/SIP-med-LLM/SIP-jmed-llm-2-8x13b-OP-instruct.
%R 10.63317/47uvbxqph5ph
%U https://aclanthology.org/2026.lrec-1.817/
%U https://doi.org/10.63317/47uvbxqph5ph
%P 10405-10423
Markdown (Informal)
[Building Effective Japanese Medical LLMs with an Open Recipe for Domain Adaptation through Continued Pre-training](https://aclanthology.org/2026.lrec-1.817/) (Aizawa et al., LREC 2026)
ACL
- Akiko Aizawa, Yuki Arase, Fei Cheng, Jiahao Huang, Zhiyi Huang, Junfeng Jiang, Teruhito Kanazawa, Daisuke Kawahara, Kazuma Kobayashi, Takashi Kodama, Sadao Kurohashi, Yusuke Oda, Yuma Tsuta, Zhen Wan, Zhishen Yang, and Rio Yokota. 2026. Building Effective Japanese Medical LLMs with an Open Recipe for Domain Adaptation through Continued Pre-training. In Proceedings of the Fifteenth Language Resources and Evaluation Conference, pages 10405–10423, Palma de Mallorca, Spain. ELRA Language Resource Association.