mlsanityData to 14 Oct 2026

Translation with Thought: Difficulty-Adaptive Reasoning via Reinforcement Learning for Multi-Domain Machine Translation

arXiv 2607.29287 ↗First posted 31 Jul 2026cs.CL cs.AIcited by 0

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Multi-domain machine translation (MDMT) poses a unique challenge due to varying levels of linguistic complexity across domains. Inspired by human translators' ability to adapt reasoning effort based on difficulty, we propose TwT (Translation with Thought), a resource-rational framework that learns to modulate inference between intuitive and deliberate reasoning. TwT is trained in two stages: (1) supervised fine-tuning on difficulty-aware long chain-of-thought traces distilled from DeepSeek-R1 and rewritten by GPT-4o to reflect human-like reasoning economy, and (2) reinforcement learning with a hybrid reward to optimize translation quality and reasoning efficiency. Evaluated on 15 benchmarks spanning in-domain and out-of-domain settings, as well as 3 seen and 59 unseen languages, with ablations across three backbone models, TwT-7B and TwT-14B outperform much larger SOTA reasoning models in translation quality, while reducing token usage by 32--60%. These results confirm that aligning translation behavior with cognitive principles enables robust generalization, high translation quality, and efficient reasoning in MDMT.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 27 of its 45 references that are in the map. First posted on arXiv 31 Jul 2026.

Builds on

27 of 45 references.

Closest ancestors · 6

Same direction, newest first

  1. 2025Test-Time Scaling of Reasoning Models for Machine Translation7 Oct
  2. 2025ExTrans: Multilingual Deep Reasoning Translation via Exemplar-Enhanced Reinforcement Learning19 May
  3. 2025MT-R1-Zero: Advancing LLM-based Machine Translation via R1-Zero-like Reinforcement Learning14 Apr
  4. 2025The Box is in the Pen: Evaluating Commonsense Reasoning in Neural Machine Translation5 Mar
  5. 2025Evaluating o1-Like LLMs: Unlocking Reasoning for Translation through Comprehensive Analysis17 Feb
  6. 2024DRT: Deep Reasoning Translation via Long Chain-of-Thought23 Dec

Other references · 21

  1. 2017Proximal Policy Optimization Algorithmscited by 15,215
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  3. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  4. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 2,057
  5. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,998
  6. 2020COMET: A Neural Framework for MT Evaluationcited by 1,033
  7. 2024Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translationcited by 394
  8. 2025Demystifying Long Chain-of-Thought Reasoning in LLMscited by 326
  9. 2022CometKiwi: IST-Unbabel 2022 Submission for the Quality Estimation Shared Taskcited by 207
  10. 2024Tower: An Open Multilingual Large Language Model for Translation-Related Taskscited by 206
  11. 2020Unsupervised Domain Clusters in Pretrained Language Modelscited by 160
  12. 2024Marco-o1: Towards Open Reasoning Models for Open-Ended Solutionscited by 144
9 more
  1. 2023Benchmarking Machine Translation with Cultural Awarenesscited by 53
  2. 2025Tower+: Bridging Generality and Translation Specialization in Multilingual LLMscited by 40
  3. 2024TEaR: Improving LLM-based Machine Translation with Systematic Self-Refinementcited by 39
  4. 2024TasTe: Teaching Large Language Models to Translate through Self-Reflectioncited by 27
  5. 2023Findings of the WMT 2023 Shared Task on Discourse-Level Literary Translation: A Fresh Orb in the Cosmos of LLMscited by 25
  6. 2024DUAL-REFLECT: Enhancing Large Language Models for Reflective Translation through Dual Learning Feedback Mechanismscited by 16
  7. 2024Findings of the WMT 2024 Shared Task on Discourse-Level Literary Translationcited by 15
  8. 2025SSR-Zero: Simple Self-Rewarding Reinforcement Learning for Machine Translationcited by 13
  9. 2024Large Language Model for Multi-Domain Translation: Benchmarking and Domain CoT Fine-tuningcited by 11

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025R1-T1: Fully Incentivizing Translation Capability in LLMs via Reasoning Learning
  2. 2025LLM Reasoning for Machine Translation: Synthetic Data Generation over Thinking Tokens
  3. 2025Chain-of-Thought Reasoning Improves Context-Aware Translation with Large Language Models
  4. 2026PAMT: Process-Aligned Reinforcement Learning for Multi-Domain Machine Translation
  5. 2025DeepTrans: Deep Reasoning Translation via Reinforcement Learning
  6. 2026ReflectMT: Internalizing Reflection for Efficient and High-Quality Machine Translation
  7. 2025New Trends for Modern Machine Translation with Large Reasoning Models
  8. 2025Please Translate Again: Two Simple Experiments on Whether Human-Like Reasoning Helps Translation