Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.19144 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917425290674176 |
|---|---|
| author | Li, Kunquan Zhang, Yingxue Meng, Fandong Su, Jinsong |
| author_facet | Li, Kunquan Zhang, Yingxue Meng, Fandong Su, Jinsong |
| contents | Recent years have witnessed growing interest in applying Large Reasoning Models (LRMs) to Machine Translation (MT). Existing approaches predominantly adopt a "think-first-then-translate" paradigm. Although explicit reasoning trajectories significantly enhance translation quality, they incur prohibitive inference costs and latency. To address these limitations, we propose ReflectMT, a two-stage reflection internalization algorithm for machine translation that employs a "translate-first-think-later" paradigm. Our approach develops the model's "translate-reflect-refine" capability through reinforcement learning. In the first stage, we cultivate the model's capacity for high-quality reflection and refinement, thereby enhancing its semantic comprehension and task-specific knowledge. In the second stage, we train the model to internalize the knowledge acquired during reflection. As a result, during inference, ReflectMT operates in a direct translation mode, producing high-quality translations on the first attempt without any explicit reasoning steps. Experimental results on datasets such as WMT24 demonstrate that our model's first-pass translations during inference outperform multi-step reasoning LRMs such as DeepSeek-R1 in both automatic metrics and GPT-based evaluation, achieving a 2.16-point improvement in GPT-based translation quality evaluation while reducing token consumption by 94.33%. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_19144 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ReflectMT: Internalizing Reflection for Efficient and High-Quality Machine Translation Li, Kunquan Zhang, Yingxue Meng, Fandong Su, Jinsong Computation and Language Recent years have witnessed growing interest in applying Large Reasoning Models (LRMs) to Machine Translation (MT). Existing approaches predominantly adopt a "think-first-then-translate" paradigm. Although explicit reasoning trajectories significantly enhance translation quality, they incur prohibitive inference costs and latency. To address these limitations, we propose ReflectMT, a two-stage reflection internalization algorithm for machine translation that employs a "translate-first-think-later" paradigm. Our approach develops the model's "translate-reflect-refine" capability through reinforcement learning. In the first stage, we cultivate the model's capacity for high-quality reflection and refinement, thereby enhancing its semantic comprehension and task-specific knowledge. In the second stage, we train the model to internalize the knowledge acquired during reflection. As a result, during inference, ReflectMT operates in a direct translation mode, producing high-quality translations on the first attempt without any explicit reasoning steps. Experimental results on datasets such as WMT24 demonstrate that our model's first-pass translations during inference outperform multi-step reasoning LRMs such as DeepSeek-R1 in both automatic metrics and GPT-based evaluation, achieving a 2.16-point improvement in GPT-based translation quality evaluation while reducing token consumption by 94.33%. |
| title | ReflectMT: Internalizing Reflection for Efficient and High-Quality Machine Translation |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2604.19144 |