Saved in:
Bibliographic Details
Main Authors: Ri, Ryokan, Kiyono, Shun, Takase, Sho
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2407.00454
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929501170040832
author Ri, Ryokan
Kiyono, Shun
Takase, Sho
author_facet Ri, Ryokan
Kiyono, Shun
Takase, Sho
contents Zero-shot cross-lingual transfer by fine-tuning multilingual pretrained models shows promise for low-resource languages, but often suffers from misalignment of internal representations between languages. We hypothesize that even when the model cannot generalize across languages effectively in fine-tuning, it still captures cross-lingual correspondence useful for cross-lingual transfer. We explore this hypothesis with Self-Translate-Train, a method that lets large language models (LLMs) to translate training data into the target language and fine-tunes the model on its own generated data. By demonstrating that Self-Translate-Train outperforms zero-shot transfer, we encourage further exploration of better methods to elicit cross-lingual capabilities of LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2407_00454
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Self-Translate-Train: Enhancing Cross-Lingual Transfer of Large Language Models via Inherent Capability
Ri, Ryokan
Kiyono, Shun
Takase, Sho
Computation and Language
Zero-shot cross-lingual transfer by fine-tuning multilingual pretrained models shows promise for low-resource languages, but often suffers from misalignment of internal representations between languages. We hypothesize that even when the model cannot generalize across languages effectively in fine-tuning, it still captures cross-lingual correspondence useful for cross-lingual transfer. We explore this hypothesis with Self-Translate-Train, a method that lets large language models (LLMs) to translate training data into the target language and fine-tunes the model on its own generated data. By demonstrating that Self-Translate-Train outperforms zero-shot transfer, we encourage further exploration of better methods to elicit cross-lingual capabilities of LLMs.
title Self-Translate-Train: Enhancing Cross-Lingual Transfer of Large Language Models via Inherent Capability
topic Computation and Language
url https://arxiv.org/abs/2407.00454