Toward LLMs Beyond English-Centric Development
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911687739703296 |
|---|---|
| author | Takase, Sho Honda, Ukyo |
| author_facet | Takase, Sho Honda, Ukyo |
| contents | Through an analysis of sequences generated by open-weight large language models (LLMs), we demonstrate that LLMs are heavily biased toward English. While continual pre-training is commonly used to adapt LLMs to a target language, we show that it does not offer a cost advantage over training from scratch, even for improving cultural understanding in the target language. These findings suggest that dedicated per-language investment may become increasingly important for future LLM development, rather than relying primarily on the expansion of English-centric resources. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_15613 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Toward LLMs Beyond English-Centric Development Takase, Sho Honda, Ukyo Computation and Language Through an analysis of sequences generated by open-weight large language models (LLMs), we demonstrate that LLMs are heavily biased toward English. While continual pre-training is commonly used to adapt LLMs to a target language, we show that it does not offer a cost advantage over training from scratch, even for improving cultural understanding in the target language. These findings suggest that dedicated per-language investment may become increasingly important for future LLM development, rather than relying primarily on the expansion of English-centric resources. |
| title | Toward LLMs Beyond English-Centric Development |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2605.15613 |