Unlocking Fine-Grained Translation Quality Estimation in LRMs through Synergistically Evolving Implicit and Explicit Reasoning
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914617917177856 |
|---|---|
| author | Dang, Renfei Wang, Xinye Lai, Zhejian Xu, Weilu Tao, Shimin Wei, Daimeng Zhang, Min Huang, Shujian |
| author_facet | Dang, Renfei Wang, Xinye Lai, Zhejian Xu, Weilu Tao, Shimin Wei, Daimeng Zhang, Min Huang, Shujian |
| contents | Large Reasoning Models (LRMs) still struggle with fine-grained translation quality estimation (QE), even with long reasoning chains. We argue that LRMs already possess strong multilingual capabilities, while the core challenge stems from the intrinsic difficulty of learning the fine-grained QE task. In this paper, we propose RIEQE (Reasoning both Implicitly and Explicitly for QE), a simple two-stage training framework that enables the co-evolution of implicit (layer-wise) and explicit (token-wise) reasoning capabilities. To make implicit reasoning feasible, we first decompose the complex QE task into straightforward subtasks. Based on this, our two-stage approach applies: (1) NonThinking-SFT, Supervised Fine-Tuning (SFT) without reasoning chains to directly boost the model's implicit reasoning tendency and capability; and (2) Thinking-RLVR, standard Reinforcement Learning with Verifiable Reward (RLVR) to subsequently strengthen explicit reasoning. Results demonstrate that implicit and explicit reasoning synergistically co-evolve under our framework. On the WMT test sets, RIEQE based on Qwen3-4B-Thinking-2507 surpasses all baselines in explicit reasoning performance, while its implicit reasoning capability is also comparable to the best current encoder-based models. We further provide evidence for the synergistic collaboration between implicit and explicit reasoning, showing how they mutually benefit each other. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_31378 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Unlocking Fine-Grained Translation Quality Estimation in LRMs through Synergistically Evolving Implicit and Explicit Reasoning Dang, Renfei Wang, Xinye Lai, Zhejian Xu, Weilu Tao, Shimin Wei, Daimeng Zhang, Min Huang, Shujian Computation and Language Large Reasoning Models (LRMs) still struggle with fine-grained translation quality estimation (QE), even with long reasoning chains. We argue that LRMs already possess strong multilingual capabilities, while the core challenge stems from the intrinsic difficulty of learning the fine-grained QE task. In this paper, we propose RIEQE (Reasoning both Implicitly and Explicitly for QE), a simple two-stage training framework that enables the co-evolution of implicit (layer-wise) and explicit (token-wise) reasoning capabilities. To make implicit reasoning feasible, we first decompose the complex QE task into straightforward subtasks. Based on this, our two-stage approach applies: (1) NonThinking-SFT, Supervised Fine-Tuning (SFT) without reasoning chains to directly boost the model's implicit reasoning tendency and capability; and (2) Thinking-RLVR, standard Reinforcement Learning with Verifiable Reward (RLVR) to subsequently strengthen explicit reasoning. Results demonstrate that implicit and explicit reasoning synergistically co-evolve under our framework. On the WMT test sets, RIEQE based on Qwen3-4B-Thinking-2507 surpasses all baselines in explicit reasoning performance, while its implicit reasoning capability is also comparable to the best current encoder-based models. We further provide evidence for the synergistic collaboration between implicit and explicit reasoning, showing how they mutually benefit each other. |
| title | Unlocking Fine-Grained Translation Quality Estimation in LRMs through Synergistically Evolving Implicit and Explicit Reasoning |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2605.31378 |