mlsanityData to 14 Oct 2026

Backtranslation Augmented Direct Preference Optimization for Neural Machine Translation

arXiv 2604.25702 ↗First posted 28 Apr 2026cs.CLcited by 0

On the map: LLM alignment / RLHF / preference optimisation → LLM preference optimization methods

Abstract (arXiv)

Contemporary neural machine translation (NMT) systems are almost exclusively built by training on supervised parallel data. Despite the tremendous progress achieved, these systems still exhibit persistent translation errors. This paper proposes that a post-training paradigm based on reinforcement learning (RL) can effectively rectify such mistakes. We introduce a novel framework that requires only a general text corpus and an expert translator which can be either human or an AI system to provide iterative feedback. In our experiments, we focus specifically on English-to-German translation as a representative high-resource language pair. Crucially, we implement this RL-based post-training using Direct Preference Optimization (DPO). Applying our DPO-driven framework to the gemma3-1b model yields a significant improvement in translation quality, elevating its COMET score from 0.703 to 0.747 on the English to German task. The results demonstrate that DPO offers an efficient and stable pathway for enhancing pre-trained NMT models through preference-based post-training.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in LLM preference optimization methods, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 12 of its 21 references that are in the map. First posted on arXiv 28 Apr 2026.

Builds on

12 of 21 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Improving LLMs for Machine Translation Using Synthetic Preference Data20 Aug
  2. 2024Cross-lingual Human-Preference Alignment for Neural Machine Translation with Direct Quality Optimization26 Sep
  3. 2024SimPO: Simple Preference Optimization with a Reference-Free Reward23 May
  4. 2024ORPO: Monolithic Preference Optimization without Reference Model12 Mar

Other references · 8

  1. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  2. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  3. 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,418
  4. 2025Gemma 3 Technical Reportcited by 1,824
  5. 2020COMET: A Neural Framework for MT Evaluationcited by 1,033
  6. 2024DeepSeek LLM: Scaling Open-Source Language Models with Longtermismcited by 642
  7. 2019Neural Machine Translation: A Review and Surveycited by 100
  8. 2026PEGRL: Improving Machine Translation by Post-Editing Guided Reinforcement Learningcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Advancing Translation Preference Modeling with RLHF: A Step Towards Cost-Effective Solution
  2. 2023Direct Preference Optimization for Neural Machine Translation with Minimum Bayes Risk Decoding
  3. 2025RL from Teacher-Model Refinement: Gradual Imitation Learning for Machine Translation
  4. 2026Rewrite to Translate, Translate to Reward: Reinforcement Learning for Source Rewriting in Machine Translation
  5. 2025CRPO: Confidence-Reward Driven Preference Optimization for Machine Translation
  6. 2025RIVAL: Reinforcement Learning with Iterative and Adversarial Optimization for Machine Translation
  7. 2026Translate-R1: Cost-Aware Translation Tool Use via Reinforcement Learning
  8. 2024Optimizing the Training Schedule of Multilingual NMT using Reinforcement Learning