{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,2]],"date-time":"2026-09-02T20:22:48Z","timestamp":1788380568936,"version":"build-2803163510"},"reference-count":56,"publisher":"MIT Press","license":[{"start":{"date-parts":[[2026,9,2]],"date-time":"2026-09-02T00:00:00Z","timestamp":1788307200000},"content-version":"vor","delay-in-days":244,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["direct.mit.edu"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2026,8,28]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>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.<\/jats:p>","DOI":"10.1162\/tacl.a.793","type":"journal-article","created":{"date-parts":[[2026,9,2]],"date-time":"2026-09-02T20:01:32Z","timestamp":1788379292000},"page":"2103-2119","update-policy":"https:\/\/doi.org\/10.1162\/mitpressjournals.corrections.policy","source":"Crossref","is-referenced-by-count":0,"title":["R1-T1: Fully Incentivizing Translation Capability in LLMs via Reasoning Learning"],"prefix":"10.1162","volume":"14","author":[{"given":"Minggui","family":"He","sequence":"first","affiliation":[{"name":"Waseda University, Fukuoka, Japan"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yilun","family":"Liu","sequence":"additional","affiliation":[{"name":"Huawei Technologies Co., Ltd., China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shimin","family":"Tao","sequence":"additional","affiliation":[{"name":"Huawei Technologies Co., Ltd., China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hongyong","family":"Zeng","sequence":"additional","affiliation":[{"name":"Xiamen University, Xiamen, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jian","family":"Zhang","sequence":"additional","affiliation":[{"name":"Waseda University, Fukuoka, Japan"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yuanchang","family":"Luo","sequence":"additional","affiliation":[{"name":"Huawei Technologies Co., Ltd., China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Li","family":"Zhang","sequence":"additional","affiliation":[{"name":"Huawei Technologies Co., Ltd., China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Daimeng","family":"Wei","sequence":"additional","affiliation":[{"name":"Huawei Technologies Co., Ltd., China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Weibin","family":"Meng","sequence":"additional","affiliation":[{"name":"Huawei Technologies Co., Ltd., China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Osamu","family":"Yoshie","sequence":"additional","affiliation":[{"name":"Waseda University, Fukuoka, Japan"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"281","published-online":{"date-parts":[[2026,8,28]]},"reference":[{"key":"2026090216011793400_bib1","doi-asserted-by":"crossref","first-page":"14438","DOI":"10.18653\/v1\/2024.emnlp-main.802","article-title":"Can automatic metrics assess high-quality translations?","volume-title":"Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing","author":"Agrawal","year":"2024"},{"key":"2026090216011793400_bib2","volume-title":"Contextualizing translation theories: Aspects of Arabic\u2013English interlingual communication","author":"Almanna","year":"2015"},{"key":"2026090216011793400_bib3","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2020.acl-main.417","article-title":"Paracrawl: Web-scale acquisition of parallel corpora","volume-title":"Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics","author":"Ba\u00f1\u00f3n","year":"2020"},{"key":"2026090216011793400_bib4","doi-asserted-by":"crossref","first-page":"1","DOI":"10.18653\/v1\/W19-5301","article-title":"Findings of the 2019 conference on machine translation (WMT19)","volume-title":"Proceedings of the Fourth Conference on Machine Translation (Volume 2: Shared Task Papers, Day 1)","author":"Barrault","year":"2019"},{"key":"2026090216011793400_bib5","doi-asserted-by":"crossref","first-page":"272","DOI":"10.18653\/v1\/W18-6401","article-title":"Findings of the 2018 conference on machine translation (WMT18)","volume-title":"Proceedings of the Third Conference on Machine Translation: Shared Task Papers","author":"Bojar","year":"2018"},{"key":"2026090216011793400_bib6","doi-asserted-by":"crossref","first-page":"73","DOI":"10.18653\/v1\/2024.arabicnlp-1.7","article-title":"Improving language models trained on translated data with continual pre-training and dictionary learning analysis","volume-title":"Proceedings of the Second Arabic Natural Language Processing Conference","author":"Boughorbel","year":"2024"},{"key":"2026090216011793400_bib7","first-page":"56","article-title":"Translation as interpretation","volume-title":"Translation studies: Perspectives on an emerging discipline","author":"B\u00fchler","year":"2002"},{"issue":"2","key":"2026090216011793400_bib8","doi-asserted-by":"crossref","first-page":"365","DOI":"10.1080\/0907676X.2021.2007272","article-title":"The use of context in multiword-term translation","volume":"31","author":"Cabezas-Garc\u00eda","year":"2023","journal-title":"Perspectives"},{"key":"2026090216011793400_bib9","first-page":"10818","article-title":"SFT memorizes, RL generalizes: A comparative study of foundation model post-training","volume-title":"International Conference on Machine Learning","author":"Chu","year":"2025"},{"key":"2026090216011793400_bib10","first-page":"3330","article-title":"Domaino1s: Guiding LLM reasoning for explainable answers in high-stakes domains","volume-title":"Findings of the Association for Computational Linguistics: ACL 2025","author":"Chu","year":"2025"},{"key":"2026090216011793400_bib11","article-title":"DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning","volume-title":"arXiv preprint arXiv:2501.12948v1","author":"DeepSeek-AI","year":"2025"},{"key":"2026090216011793400_bib12","doi-asserted-by":"crossref","first-page":"3922","DOI":"10.18653\/v1\/2025.findings-naacl.218","article-title":"TEaR: Improving LLM-based machine translation with systematic self-refinement","volume-title":"Findings of the Association for Computational Linguistics: NAACL 2025","author":"Feng","year":"2025"},{"key":"2026090216011793400_bib13","doi-asserted-by":"crossref","first-page":"1460","DOI":"10.1162\/tacl_a_00437","article-title":"Experts, errors, and context: A large-scale study of human evaluation for machine translation","volume":"9","author":"Freitag","year":"2021","journal-title":"Transactions of the Association for Computational Linguistics"},{"key":"2026090216011793400_bib14","article-title":"Continual learning for neural machine translation","volume-title":"Proceedings of the 28th International Conference on Computational Linguistics","author":"Fujii","year":"2020"},{"key":"2026090216011793400_bib15","doi-asserted-by":"crossref","first-page":"522","DOI":"10.1162\/tacl_a_00474","article-title":"The FLORES-101 evaluation benchmark for low-resource and multilingual machine translation","volume":"10","author":"Goyal","year":"2022","journal-title":"Transactions of the Association for Computational Linguistics"},{"key":"2026090216011793400_bib16","doi-asserted-by":"crossref","first-page":"979","DOI":"10.1162\/tacl_a_00683","article-title":"xCOMET: Transparent machine translation evaluation through fine-grained error detection","volume":"12","author":"Guerreiro","year":"2024","journal-title":"Transactions of the Association for Computational Linguistics"},{"key":"2026090216011793400_bib17","doi-asserted-by":"crossref","first-page":"3662","DOI":"10.18653\/v1\/2020.findings-emnlp.327","article-title":"The box is in the pen: Evaluating commonsense reasoning in neural machine translation","volume-title":"Findings of the Association for Computational Linguistics: EMNLP 2020","author":"He","year":"2020"},{"key":"2026090216011793400_bib18","article-title":"GPT-4o system card","volume-title":"arXiv preprint arXiv:2410.21276v1","author":"Hurst","year":"2024"},{"key":"2026090216011793400_bib19","article-title":"OpenAI o1 system card","volume-title":"arXiv preprint arXiv:2412.16720v1","author":"Jaech","year":"2024"},{"key":"2026090216011793400_bib20","first-page":"73","article-title":"Gradable chatgpt translation evaluation","volume":"72","author":"Jiao","year":"2024","journal-title":"Procesamiento del Lenguaje Natural"},{"key":"2026090216011793400_bib21","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2023.findings-emnlp.1001","article-title":"Parrot: Translating during chat using large language models tuned with human translation and feedback","volume-title":"Findings of the Association for Computational Linguistics: EMNLP 2023","author":"Jiao","year":"2023"},{"key":"2026090216011793400_bib22","article-title":"Is ChatGPT a good translator? yes with GPT-4 as the engine","volume-title":"arXiv preprint arXiv:2301.08745v4","author":"Jiao","year":"2023"},{"key":"2026090216011793400_bib23","doi-asserted-by":"crossref","first-page":"756","DOI":"10.18653\/v1\/2023.wmt-1.63","article-title":"MetricX-23: The google submission to the WMT 2023 metrics shared task","volume-title":"Proceedings of the Eighth Conference on Machine Translation","author":"Juraska","year":"2023"},{"key":"2026090216011793400_bib24","doi-asserted-by":"crossref","first-page":"78","DOI":"10.18653\/v1\/2025.mrl-main.6","article-title":"Understand, solve and translate: Bridging the multilingual mathematical reasoning gap","volume-title":"Proceedings of the 5th Workshop on Multilingual Representation Learning (MRL 2025)","author":"Ko","year":"2025"},{"key":"2026090216011793400_bib25","doi-asserted-by":"crossref","first-page":"1","DOI":"10.18653\/v1\/2024.wmt-1.1","article-title":"Findings of the WMT24 general machine translation shared task: The LLM era is here but MT is not solved yet","volume-title":"Proceedings of the Ninth Conference on Machine Translation","author":"Kocmi","year":"2024"},{"key":"2026090216011793400_bib26","article-title":"Preliminary WMT24 ranking of general MT systems and LLMs","volume-title":"arXiv preprint arXiv:2407.19884v1","author":"Kocmi","year":"2024"},{"key":"2026090216011793400_bib27","doi-asserted-by":"crossref","DOI":"10.1163\/9789004495746","volume-title":"Toward a science of translating: with special reference to principles and procedures involved in Bible translating","author":"Nida","year":"1964"},{"key":"2026090216011793400_bib28","article-title":"No language left behind: Scaling human-centered machine translation","volume-title":"arXiv preprint arXiv:2207.04672v3","author":"Team","year":"2022"},{"issue":"3","key":"2026090216011793400_bib29","first-page":"8","article-title":"Translation procedures, strategies and methods","volume":"11","author":"Ordudari","year":"2007","journal-title":"Translation journal"},{"key":"2026090216011793400_bib30","doi-asserted-by":"crossref","first-page":"27730","DOI":"10.52202\/068431-2011","article-title":"Training language models to follow instructions with human feedback","volume":"35","author":"Ouyang","year":"2022","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2026090216011793400_bib31","doi-asserted-by":"crossref","first-page":"5622","DOI":"10.18653\/v1\/2023.findings-emnlp.373","article-title":"Towards making the most of ChatGPT for machine translation","volume-title":"Findings of the Association for Computational Linguistics: EMNLP 2023","author":"Peng","year":"2023"},{"key":"2026090216011793400_bib32","article-title":"o1 replication journey: A strategic progress report\u2013part 1","volume-title":"arXiv preprint arXiv:2410.18982v1","author":"Qin","year":"2024"},{"key":"2026090216011793400_bib33","doi-asserted-by":"crossref","first-page":"578","DOI":"10.18653\/v1\/2022.wmt-1.52","article-title":"COMET-22: Unbabelist 2022 submission for the metrics shared task","volume-title":"Proceedings of the Seventh Conference on Machine Translation (WMT)","author":"Rei","year":"2022"},{"key":"2026090216011793400_bib34","doi-asserted-by":"crossref","first-page":"2685","DOI":"10.18653\/v1\/2020.emnlp-main.213","article-title":"COMET: A neural framework for MT evaluation","volume-title":"Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)","author":"Rei","year":"2020"},{"key":"2026090216011793400_bib35","doi-asserted-by":"crossref","first-page":"141","DOI":"10.1075\/hts.3.rel4","article-title":"Relay translation","volume":"3","author":"Ringmar","year":"2012","journal-title":"Handbook of translation studies"},{"key":"2026090216011793400_bib36","article-title":"Proximal policy optimization algorithms","volume-title":"arXiv preprint arXiv:1707.06347v2","author":"Schulman","year":"2017"},{"key":"2026090216011793400_bib37","article-title":"Do multilingual LLMs think in english?","volume-title":"ICLR 2025 Workshop on Building Trust in Language Models and Applications","author":"Schut","year":"2025"},{"key":"2026090216011793400_bib38","article-title":"DeepSeekMath: Pushing the limits of mathematical reasoning in open language models","volume-title":"arXiv preprint arXiv:2402.03300v3","author":"Shao","year":"2024"},{"key":"2026090216011793400_bib39","article-title":"VLM-R1: A stable and generalizable R1-style large vision-language model","volume-title":"arXiv preprint arXiv:2504.07615v2","author":"Shen","year":"2025"},{"key":"2026090216011793400_bib40","first-page":"127","article-title":"Round-trip translation: What is it good for?","volume-title":"Proceedings of the Australasian Language Technology Workshop 2005","author":"Somers","year":"2005"},{"issue":"1","key":"2026090216011793400_bib41","doi-asserted-by":"crossref","DOI":"10.32604\/cmc.2024.054673","article-title":"Lkmt: Linguistics knowledge-driven multi-task neural machine translation for urdu and english","volume":"81","author":"Ul Hassan","year":"2024","journal-title":"Computers, Materials & Continua"},{"key":"2026090216011793400_bib42","doi-asserted-by":"crossref","first-page":"6770","DOI":"10.18653\/v1\/2025.findings-acl.351","article-title":"DRT: Deep reasoning translation via long chain-of-thought","volume-title":"Findings of the Association for Computational Linguistics: ACL 2025","author":"Wang","year":"2025"},{"key":"2026090216011793400_bib43","doi-asserted-by":"crossref","first-page":"5858","DOI":"10.18653\/v1\/2025.findings-emnlp.313","article-title":"Retrieval-augmented machine translation with unstructured knowledge","volume-title":"Findings of the Association for Computational Linguistics: EMNLP 2025","author":"Wang","year":"2025"},{"key":"2026090216011793400_bib44","doi-asserted-by":"crossref","first-page":"6144","DOI":"10.18653\/v1\/2024.acl-long.333","article-title":"TasTe: Teaching large language models to translate through self-reflection","volume-title":"Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","author":"Wang","year":"2024"},{"key":"2026090216011793400_bib45","doi-asserted-by":"crossref","first-page":"24824","DOI":"10.52202\/068431-1800","article-title":"Chain-of-thought prompting elicits reasoning in large language models","volume":"35","author":"Wei","year":"2022","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2026090216011793400_bib46","article-title":"Adapting large language models for document-level machine translation","volume-title":"arXiv preprint arXiv:2401.06468v4","author":"Wu","year":"2024"},{"key":"2026090216011793400_bib47","article-title":"A paradigm shift in machine translation: Boosting translation performance of large language models","volume-title":"The Twelfth International Conference on Learning Representations","author":"Xu","year":"2024"},{"key":"2026090216011793400_bib48","article-title":"X-ALMA: Plug & play modules and adaptive rejection for quality translation at scale","volume-title":"Proceedings of the Thirteenth International Conference on Learning Representations","author":"Xu","year":"2025"},{"key":"2026090216011793400_bib49","article-title":"Qwen2.5 technical report","volume-title":"arXiv preprint arXiv:2412.15115v2","author":"Yang","year":"2024"},{"key":"2026090216011793400_bib50","doi-asserted-by":"crossref","first-page":"13078","DOI":"10.18653\/v1\/2024.findings-emnlp.765","article-title":"Benchmarking machine translation with cultural awareness","volume-title":"Findings of the Association for Computational Linguistics: EMNLP 2024","author":"Yao","year":"2024"},{"key":"2026090216011793400_bib51","doi-asserted-by":"crossref","first-page":"6029","DOI":"10.18653\/v1\/2023.acl-long.332","article-title":"Understanding and improving the robustness of terminology constraints in neural machine translation","volume-title":"Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","author":"Zhang","year":"2023"},{"key":"2026090216011793400_bib52","doi-asserted-by":"crossref","first-page":"468","DOI":"10.18653\/v1\/2023.wmt-1.43","article-title":"Machine translation with large language models: Prompting, few-shot learning, and fine-tuning with QLoRA","volume-title":"Proceedings of the Eighth Conference on Machine Translation","author":"Zhang","year":"2023"},{"key":"2026090216011793400_bib53","article-title":"o1-Coder: an o1 replication for coding","volume-title":"arXiv preprint arXiv:2412.00154v2","author":"Zhang","year":"2024"},{"key":"2026090216011793400_bib54","doi-asserted-by":"crossref","first-page":"7025","DOI":"10.18653\/v1\/2024.acl-long.379","article-title":"PLUG: Leveraging pivot language in cross-lingual instruction tuning","volume-title":"Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","author":"Zhang","year":"2024"},{"key":"2026090216011793400_bib55","article-title":"Marco-o1: Towards open reasoning models for open-ended solutions","volume-title":"arXiv preprint arXiv:2411.14405v2","author":"Zhao","year":"2024"},{"issue":"3","key":"2026090216011793400_bib56","doi-asserted-by":"crossref","first-page":"104078","DOI":"10.1016\/j.ipm.2025.104078","article-title":"Overcoming language barriers via machine translation with sparse mixture-of-experts fusion of large language models","volume":"62","author":"Zhu","year":"2025","journal-title":"Information Processing & Management"}],"container-title":["Transactions of the Association for Computational Linguistics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/direct.mit.edu\/tacl\/article-pdf\/doi\/10.1162\/TACL.a.793\/2624695\/tacl.a.793.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/direct.mit.edu\/tacl\/article-pdf\/doi\/10.1162\/TACL.a.793\/2624695\/tacl.a.793.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,2]],"date-time":"2026-09-02T20:01:46Z","timestamp":1788379306000},"score":1,"resource":{"primary":{"URL":"https:\/\/direct.mit.edu\/tacl\/article\/doi\/10.1162\/TACL.a.793\/138667\/R1-T1-Fully-Incentivizing-Translation-Capability"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":56,"URL":"https:\/\/doi.org\/10.1162\/tacl.a.793","relation":{},"ISSN":["2307-387X"],"issn-type":[{"value":"2307-387X","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2026]]},"published":{"date-parts":[[2026]]}}}