Lost in Literalism: How Supervised Training Shapes Translationese in LLMs
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866909527705649152 |
|---|---|
| author | Li, Yafu Zhang, Ronghao Wang, Zhilin Zhang, Huajian Cui, Leyang Yin, Yongjing Xiao, Tong Zhang, Yue |
| author_facet | Li, Yafu Zhang, Ronghao Wang, Zhilin Zhang, Huajian Cui, Leyang Yin, Yongjing Xiao, Tong Zhang, Yue |
| contents | Large language models (LLMs) have achieved remarkable success in machine translation, demonstrating impressive performance across diverse languages. However, translationese, characterized by overly literal and unnatural translations, remains a persistent challenge in LLM-based translation systems. Despite their pre-training on vast corpora of natural utterances, LLMs exhibit translationese errors and generate unexpected unnatural translations, stemming from biases introduced during supervised fine-tuning (SFT). In this work, we systematically evaluate the prevalence of translationese in LLM-generated translations and investigate its roots during supervised training. We introduce methods to mitigate these biases, including polishing golden references and filtering unnatural training instances. Empirical evaluations demonstrate that these approaches significantly reduce translationese while improving translation naturalness, validated by human evaluations and automatic metrics. Our findings highlight the need for training-aware adjustments to optimize LLM translation outputs, paving the way for more fluent and target-language-consistent translations. We release the data and code at https://github.com/yafuly/LLM_Translationese. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_04369 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Lost in Literalism: How Supervised Training Shapes Translationese in LLMs Li, Yafu Zhang, Ronghao Wang, Zhilin Zhang, Huajian Cui, Leyang Yin, Yongjing Xiao, Tong Zhang, Yue Computation and Language Large language models (LLMs) have achieved remarkable success in machine translation, demonstrating impressive performance across diverse languages. However, translationese, characterized by overly literal and unnatural translations, remains a persistent challenge in LLM-based translation systems. Despite their pre-training on vast corpora of natural utterances, LLMs exhibit translationese errors and generate unexpected unnatural translations, stemming from biases introduced during supervised fine-tuning (SFT). In this work, we systematically evaluate the prevalence of translationese in LLM-generated translations and investigate its roots during supervised training. We introduce methods to mitigate these biases, including polishing golden references and filtering unnatural training instances. Empirical evaluations demonstrate that these approaches significantly reduce translationese while improving translation naturalness, validated by human evaluations and automatic metrics. Our findings highlight the need for training-aware adjustments to optimize LLM translation outputs, paving the way for more fluent and target-language-consistent translations. We release the data and code at https://github.com/yafuly/LLM_Translationese. |
| title | Lost in Literalism: How Supervised Training Shapes Translationese in LLMs |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2503.04369 |