Backtranslation Augmented Direct Preference Optimization for Neural Machine Translation
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917550055489536 |
|---|---|
| author | Ghassabi, Mehrdad Rajabi, Spehr Kashani, Hamidreza Baradaran Hakim, Sadra Keivandarian, Mahshid Bahnamiri, Amirhossein Jahani |
| author_facet | Ghassabi, Mehrdad Rajabi, Spehr Kashani, Hamidreza Baradaran Hakim, Sadra Keivandarian, Mahshid Bahnamiri, Amirhossein Jahani |
| contents | 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 it's 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_25702 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Backtranslation Augmented Direct Preference Optimization for Neural Machine Translation Ghassabi, Mehrdad Rajabi, Spehr Kashani, Hamidreza Baradaran Hakim, Sadra Keivandarian, Mahshid Bahnamiri, Amirhossein Jahani Computation and Language 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 it's 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. |
| title | Backtranslation Augmented Direct Preference Optimization for Neural Machine Translation |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2604.25702 |