@article{he-etal-2026-r1,
title = "R1-T1: Fully Incentivizing Translation Capability in {LLM}s via Reasoning Learning",
author = "He, Minggui and
Liu, Yilun and
Tao, Shimin and
Zeng, Hongyong and
Zhang, Jian and
Luo, Yuanchang and
Zhang, Li and
Wei, Daimeng and
Meng, Weibin and
Yoshie, Osamu",
journal = "Transactions of the Association for Computational Linguistics",
volume = "14",
year = "2026",
address = "Cambridge, MA",
publisher = "MIT Press",
url = "https://aclanthology.org/2026.tacl-1.95/",
doi = "10.1162/tacl.a.793",
pages = "2103--2119",
abstract = "Despite recent breakthroughs in reasoning-enhanced large language models (LLMs), incorporating inference-time reasoning into application tasks such as machine translation (MT), where human translators naturally employ structured, multi-layered reasoning chain-of-thoughts (CoTs), is yet un-derexplored. Existing methods either design a fixed CoT tailored for a specific MT sub-task (e.g., literature translation), or rely on synthesizing CoTs unaligned with humans and supervised fine-tuning (SFT) prone to overfitting, limiting their adaptability to diverse translation scenarios. This paper introduces R1-Translator (R1-T1), a novel framework to achieve inference-time reasoning for general MT via reinforcement learning (RL) with human-aligned CoTs comprising six common patterns. Our approach pioneers three innovations: (1) verifying reasoning-based translation in various MT scenarios (e.g., multilingual MT, domain MT) unseen from the training phase; (2) formalizing six expert-curated CoT templates that mirror hybrid human strategies like context-aware paraphrasing and round-trip translation; and (3) enabling more flexible CoTs through an RL stage after cold-start. Both human and automatic evaluation results indicate a steady translation quality improvement in a total of 10+ languages and 40+ translation directions on Flores-101 test set and four domain-specific MT tasks, especially on the languages unseen from training."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="he-etal-2026-r1">
<titleInfo>
<title>R1-T1: Fully Incentivizing Translation Capability in LLMs via Reasoning Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Minggui</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yilun</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shimin</namePart>
<namePart type="family">Tao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hongyong</namePart>
<namePart type="family">Zeng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jian</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuanchang</namePart>
<namePart type="family">Luo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Li</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Daimeng</namePart>
<namePart type="family">Wei</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weibin</namePart>
<namePart type="family">Meng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Osamu</namePart>
<namePart type="family">Yoshie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<genre authority="bibutilsgt">journal article</genre>
<relatedItem type="host">
<titleInfo>
<title>Transactions of the Association for Computational Linguistics</title>
</titleInfo>
<originInfo>
<issuance>continuing</issuance>
<publisher>MIT Press</publisher>
<place>
<placeTerm type="text">Cambridge, MA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">periodical</genre>
<genre authority="bibutilsgt">academic journal</genre>
</relatedItem>
<abstract>Despite recent breakthroughs in reasoning-enhanced large language models (LLMs), incorporating inference-time reasoning into application tasks such as machine translation (MT), where human translators naturally employ structured, multi-layered reasoning chain-of-thoughts (CoTs), is yet un-derexplored. Existing methods either design a fixed CoT tailored for a specific MT sub-task (e.g., literature translation), or rely on synthesizing CoTs unaligned with humans and supervised fine-tuning (SFT) prone to overfitting, limiting their adaptability to diverse translation scenarios. This paper introduces R1-Translator (R1-T1), a novel framework to achieve inference-time reasoning for general MT via reinforcement learning (RL) with human-aligned CoTs comprising six common patterns. Our approach pioneers three innovations: (1) verifying reasoning-based translation in various MT scenarios (e.g., multilingual MT, domain MT) unseen from the training phase; (2) formalizing six expert-curated CoT templates that mirror hybrid human strategies like context-aware paraphrasing and round-trip translation; and (3) enabling more flexible CoTs through an RL stage after cold-start. Both human and automatic evaluation results indicate a steady translation quality improvement in a total of 10+ languages and 40+ translation directions on Flores-101 test set and four domain-specific MT tasks, especially on the languages unseen from training.</abstract>
<identifier type="citekey">he-etal-2026-r1</identifier>
<identifier type="doi">10.1162/tacl.a.793</identifier>
<location>
<url>https://aclanthology.org/2026.tacl-1.95/</url>
</location>
<part>
<date>2026</date>
<detail type="volume"><number>14</number></detail>
<extent unit="page">
<start>2103</start>
<end>2119</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Journal Article
%T R1-T1: Fully Incentivizing Translation Capability in LLMs via Reasoning Learning
%A He, Minggui
%A Liu, Yilun
%A Tao, Shimin
%A Zeng, Hongyong
%A Zhang, Jian
%A Luo, Yuanchang
%A Zhang, Li
%A Wei, Daimeng
%A Meng, Weibin
%A Yoshie, Osamu
%J Transactions of the Association for Computational Linguistics
%D 2026
%V 14
%I MIT Press
%C Cambridge, MA
%F he-etal-2026-r1
%X Despite recent breakthroughs in reasoning-enhanced large language models (LLMs), incorporating inference-time reasoning into application tasks such as machine translation (MT), where human translators naturally employ structured, multi-layered reasoning chain-of-thoughts (CoTs), is yet un-derexplored. Existing methods either design a fixed CoT tailored for a specific MT sub-task (e.g., literature translation), or rely on synthesizing CoTs unaligned with humans and supervised fine-tuning (SFT) prone to overfitting, limiting their adaptability to diverse translation scenarios. This paper introduces R1-Translator (R1-T1), a novel framework to achieve inference-time reasoning for general MT via reinforcement learning (RL) with human-aligned CoTs comprising six common patterns. Our approach pioneers three innovations: (1) verifying reasoning-based translation in various MT scenarios (e.g., multilingual MT, domain MT) unseen from the training phase; (2) formalizing six expert-curated CoT templates that mirror hybrid human strategies like context-aware paraphrasing and round-trip translation; and (3) enabling more flexible CoTs through an RL stage after cold-start. Both human and automatic evaluation results indicate a steady translation quality improvement in a total of 10+ languages and 40+ translation directions on Flores-101 test set and four domain-specific MT tasks, especially on the languages unseen from training.
%R 10.1162/tacl.a.793
%U https://aclanthology.org/2026.tacl-1.95/
%U https://doi.org/10.1162/tacl.a.793
%P 2103-2119
Markdown (Informal)
[R1-T1: Fully Incentivizing Translation Capability in LLMs via Reasoning Learning](https://aclanthology.org/2026.tacl-1.95/) (He et al., TACL 2026)
ACL
- Minggui He, Yilun Liu, Shimin Tao, Hongyong Zeng, Jian Zhang, Yuanchang Luo, Li Zhang, Daimeng Wei, Weibin Meng, and Osamu Yoshie. 2026. R1-T1: Fully Incentivizing Translation Capability in LLMs via Reasoning Learning. Transactions of the Association for Computational Linguistics, 14:2103–2119.