Tag-Aware Structured Text Translation: Towards a Systematic Understanding

A new arXiv paper (2609.29131v1) proposes a three-level framework for tag-aware structured text translation that combines a hybrid data synthesis strategy (Hy-LST), multi-task supervised fine-tuning across four sub-tasks, and three complementary reward functions under group relative policy optimization. Experiments across six language directions (en2zh, en2ja, en2de, en2fr, en2ru, de2fr) show each level contributes measurable improvements and the complete system significantly outperforms existing methods, with joint optimization of fluency, tag fidelity, and tag-scoped translation quality consistently beating single-reward alternatives.

arXiv:2609.29131v1 Announce Type: new Abstract: Internet texts are replete with format tags that carry structural, semantic, and functional meaning. Current large language model LLM -based translation systems struggle to balance translation fluency with tag fidelity when processing tagged text. We argue that resolving this tension requires a systematic approach at three interconnected levels: data synthesis, capability building, and multi-objective alignment. At the data level, we identify and formalize a fundamental trade-off between structural tag diversity and translation naturalness in synthetic data generation; existing methods optimize for one at the expense of the other. We propose a hybrid synthesis strategy Hy-LST combining LLM-based synthesis tag method and Two-Stage LLM-based synthesis tag method to produce both diverse and natural tagged data. At the capability level, we decompose tag-aware translation into four sub-tasks of increasing difficulty in a multi-task supervised fine-tuning framework, enabling targeted capability acquisition and knowledge transfer. At the alignment level, we design three complementary reward functions under a group relative policy optimization framework, each targeting a distinct objective fluency, tag fidelity, and tag-scoped translation quality , and show that joint optimization consistently outperforms single-reward alternatives. Experiments on six language directions en2zh, en2ja, en2de, en2fr, en2ru, de2fr demonstrate that each level contributes measurable improvements, and the complete system significantly outperforms existing methods. Qualitative analysis reveals specific error patterns and their mitigation after training with our method.