mlsanityData to 14 Oct 2026

Two Intermediate Translations Are Better Than One: Fine-tuning LLMs for Document-level Translation Refinement

arXiv 2504.05614 ↗First posted 8 Apr 2025cs.CLcited by 5

Not on the map. Why?

Abstract (arXiv)

Recent research has shown that large language models (LLMs) can enhance translation quality through self-refinement. In this paper, we build on this idea by extending the refinement from sentence-level to document-level translation, specifically focusing on document-to-document (Doc2Doc) translation refinement. Since sentence-to-sentence (Sent2Sent) and Doc2Doc translation address different aspects of the translation process, we propose fine-tuning LLMs for translation refinement using two intermediate translations, combining the strengths of both Sent2Sent and Doc2Doc. Additionally, recognizing that the quality of intermediate translations varies, we introduce an enhanced fine-tuning method with quality awareness that assigns lower weights to easier translations and higher weights to more difficult ones, enabling the model to focus on challenging translation cases. Experimental results across ten translation tasks with LLaMA-3-8B-Instruct and Mistral-Nemo-Instruct demonstrate the effectiveness of our approach.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 38 of its 53 references that are in the map. First posted on arXiv 8 Apr 2025.

Builds on

38 of 53 references.

References · 38

  1. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  2. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  3. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,455
  4. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,442
  5. 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,418
  6. 2021SimCSE: Simple Contrastive Learning of Sentence Embeddingscited by 2,173
  7. 2022No Language Left Behind: Scaling Human-Centered Machine Translationcited by 1,312
  8. 2022Contrastive Decoding: Open-ended Text Generation as Optimizationcited by 609
  9. 2024Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translationcited by 394
  10. 2022A Contrastive Framework for Neural Text Generationcited by 225
  11. 2024Tower: An Open Multilingual Large Language Model for Translation-Related Taskscited by 206
  12. 2023The unreasonable effectiveness of few-shot learning for machine translationcited by 91
26 more
  1. 2023Large language models effectively leverage document-level context for literary translation, but critical errors persistcited by 80
  2. 2022Detecting and Mitigating Hallucinations in Machine Translation: Model Internal Workings Alone Do Well, Sentence Similarity Even Bettercited by 75
  3. 2024Adapting Large Language Models for Document-Level Machine Translationcited by 59
  4. 2021BlonDe: An Automatic Evaluation Metric for Document-level Machine Translationcited by 58
  5. 2022Embarrassingly Easy Document-Level MT Metrics: How to Convert Any Pretrained Metric Into a Document-Level Metriccited by 56
  6. 2023Leveraging GPT-4 for Automatic Translation Post-Editingcited by 56
  7. 2022Towards Opening the Black Box of Neural Machine Translation: Source and Target Interpretations of the Transformercited by 45
  8. 2023LLMRefine: Pinpointing and Refining Large Language Models via Fine-Grained Actionable Feedbackcited by 39
  9. 2024TEaR: Improving LLM-based Machine Translation with Systematic Self-Refinementcited by 39
  10. 2024Translating Step-by-Step: Decomposing the Translation Process for Improved Translation Quality of Long-Form Textscited by 31
  11. 2024DelTA: An Online Document-Level Translation Agent Based on Multi-Level Memorycited by 27
  12. 2024Efficiently Exploring Large Language Models for Document-Level Machine Translation with In-context Learningcited by 23
  13. 2024Guiding Large Language Models to Post-Edit Machine Translation with Error Annotationscited by 22
  14. 2024The Fine-Tuning Paradox: Boosting Translation Quality Without Sacrificing LLM Abilitiescited by 22
  15. 2023Contextual Refinement of Translations: Large Language Models for Sentence and Document-Level Post-Editingcited by 21
  16. 2024DUAL-REFLECT: Enhancing Large Language Models for Reflective Translation through Dual Learning Feedback Mechanismscited by 16
  17. 2024Word Alignment as Preference for Machine Translationcited by 14
  18. 2024Don't Rank, Combine! Combining Machine Translation Hypotheses Using Quality Estimationcited by 13
  19. 2024Enhancing Document-level Translation of Large Language Model via Translation Mixed-instructionscited by 13
  20. 2024Ladder: A Model-Agnostic Framework Boosting LLM-based Machine Translation to the Next Levelcited by 12
  21. 2024DeMPT: Decoding-enhanced Multi-phase Prompt Tuning for Making LLMs Be Better Context-aware Translatorscited by 6
  22. 2024LLM-based Translation Inference with Iterative Bilingual Understandingcited by 6
  23. 2025Doc-Guided Sent2Sent++: A Sent2Sent++ Agent with Doc-Guided memory for Document-level Machine Translationcited by 6
  24. 2017Not All Dialogues are Created Equal: Instance Weighting for Neural Conversational Modelscited by 5
  25. 2024Prompting Large Language Models with Human Error Markings for Self-Correcting Machine Translationcited by 3
  26. 2024Refining Translations with LLMs: A Constraint-Aware Iterative Prompting Approachcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 2; 2026: 3.

5 papers.

  1. 2025M2PO: Multi-Perspective Multi-Pair Preference Optimization for Machine Translationcited by 4
  2. 2026Cross-Preference Learning for Sentence-Level and Context-Aware Machine Translationcited by 2
  3. 2026Loong: A Human-Like Long Document Translation Agent with Observe-and-Act Adaptive Context Selectioncited by 1
  4. 2026STAR : Sentence Translation Alignment Rate for Document-to-Document Machine Translationcited by 1
  5. 2025Beyond the Sentence: A Survey on Context-Aware Machine Translation with Large Language Modelscited by 0