Enhancing Non-English Capabilities of English-Centric Large Language Models through Deep Supervision Fine-Tuning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917946225328128 |
|---|---|
| author | Huo, Wenshuai Feng, Xiaocheng Huang, Yichong Fu, Chengpeng Li, Baohang Ye, Yangfan Zhang, Zhirui Tu, Dandan Tang, Duyu Lu, Yunfei Wang, Hui Qin, Bing |
| author_facet | Huo, Wenshuai Feng, Xiaocheng Huang, Yichong Fu, Chengpeng Li, Baohang Ye, Yangfan Zhang, Zhirui Tu, Dandan Tang, Duyu Lu, Yunfei Wang, Hui Qin, Bing |
| contents | Large language models (LLMs) have demonstrated significant progress in multilingual language understanding and generation. However, due to the imbalance in training data, their capabilities in non-English languages are limited. Recent studies revealed the English-pivot multilingual mechanism of LLMs, where LLMs implicitly convert non-English queries into English ones at the bottom layers and adopt English for thinking at the middle layers. However, due to the absence of explicit supervision for cross-lingual alignment in the intermediate layers of LLMs, the internal representations during these stages may become inaccurate. In this work, we introduce a deep supervision fine-tuning method (DFT) that incorporates additional supervision in the internal layers of the model to guide its workflow. Specifically, we introduce two training objectives on different layers of LLMs: one at the bottom layers to constrain the conversion of the target language into English, and another at the middle layers to constrain reasoning in English. To effectively achieve the guiding purpose, we designed two types of supervision signals: logits and feature, which represent a stricter constraint and a relatively more relaxed guidance. Our method guides the model to not only consider the final generated result when processing non-English inputs but also ensure the accuracy of internal representations. We conducted extensive experiments on typical English-centric large models, LLaMA-2 and Gemma-2, and the results on multiple multilingual datasets show that our method significantly outperforms traditional fine-tuning methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_01275 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Enhancing Non-English Capabilities of English-Centric Large Language Models through Deep Supervision Fine-Tuning Huo, Wenshuai Feng, Xiaocheng Huang, Yichong Fu, Chengpeng Li, Baohang Ye, Yangfan Zhang, Zhirui Tu, Dandan Tang, Duyu Lu, Yunfei Wang, Hui Qin, Bing Computation and Language Large language models (LLMs) have demonstrated significant progress in multilingual language understanding and generation. However, due to the imbalance in training data, their capabilities in non-English languages are limited. Recent studies revealed the English-pivot multilingual mechanism of LLMs, where LLMs implicitly convert non-English queries into English ones at the bottom layers and adopt English for thinking at the middle layers. However, due to the absence of explicit supervision for cross-lingual alignment in the intermediate layers of LLMs, the internal representations during these stages may become inaccurate. In this work, we introduce a deep supervision fine-tuning method (DFT) that incorporates additional supervision in the internal layers of the model to guide its workflow. Specifically, we introduce two training objectives on different layers of LLMs: one at the bottom layers to constrain the conversion of the target language into English, and another at the middle layers to constrain reasoning in English. To effectively achieve the guiding purpose, we designed two types of supervision signals: logits and feature, which represent a stricter constraint and a relatively more relaxed guidance. Our method guides the model to not only consider the final generated result when processing non-English inputs but also ensure the accuracy of internal representations. We conducted extensive experiments on typical English-centric large models, LLaMA-2 and Gemma-2, and the results on multiple multilingual datasets show that our method significantly outperforms traditional fine-tuning methods. |
| title | Enhancing Non-English Capabilities of English-Centric Large Language Models through Deep Supervision Fine-Tuning |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2503.01275 |