← 研踪 PaperTrailarXiv:2604.19144 · v2
ReflectMT: Internalizing Reflection for Efficient and High-Quality Machine Translation
首次提交 2026-04-21 14:48
版本更新 2026-11-02 15:16
雷达观测 2026-11-04 23:42
摘要
Recent years have witnessed growing interest in applying Large Reasoning Models (LRMs) to Machine Translation (MT). While most approaches adopt a 'pre-thinking' paradigm and benefit from explicit reasoning trajectories, they suffer from substantial inference cost and latency. To address these limitations, we propose ReflectMT, a two-stage reflection internalization framework for machine translation that employs a 'post-thinking' 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. On WMT24, ReflectMT achieves a GPT-4o-based GRF score of 86.16, outperforming DeepSeek-R1 (84.03) by 2.13 points, while using approximately 92% fewer output tokens. Independent automatic evaluation, a blinded human evaluation, and preliminary experiments on three additional translation directions provide further evidence for the framework's effectiveness.
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v2 · 2026-11-02 15:16 · 最新观测版本
ReflectMT: Internalizing Reflection for Efficient and High-Quality Machine Translation
Recent years have witnessed growing interest in applying Large Reasoning Models (LRMs) to Machine Translation (MT). While most approaches adopt a 'pre-thinking' paradigm and benefit from explicit reasoning trajectories, they suffer from substantial inference cost and latency. To address these limitations, we propose ReflectMT, a two-stage reflection internalization framework for machine translation that employs a 'post-thinking' 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. On WMT24, ReflectMT achieves a GPT-4o-based GRF score of 86.16, outperforming DeepSeek-R1 (84.03) by 2.13 points, while using approximately 92% fewer output tokens. Independent automatic evaluation, a blinded human evaluation, and preliminary experiments on three additional translation directions provide further evidence for the framework's effectiveness.
查看此版本解读 →查看此版本原文 ↗